Nrf2 Protects Pancreatic β-Cells From Oxidative and Nitrosative Stress in Diabetic Model Mice

The transcription factor Nrf2 (NF-E2–related factor 2) has been postulated to be a key regulator of antioxidant-enzyme genes (1). Nrf2 is a basic region-leucine zipper-type transcription factor belonging to the cap’n’collar family (2,3). Keap1 (Kelch-like ECH–associated protein 11) is an adaptor protein for Cullin3-based ubiquitin E3 ligase and interacts with Nrf2 (4). Keap1 represses transcriptional activity of Nrf2 by ubiquitination and subsequent degradation of Nrf2 through the proteasome pathway during unstressed conditions (5–7). However, upon exposure to electrophiles and reactive species (RS), Keap1 cysteine residues are modified (8,9), leading to disturbances of Keap1 ubiquitin ligase activity and proteasome-mediated degradation of Nrf2 (5,10,11). When Nrf2 is stabilized, it accumulates in the nucleus and heterodimerizes with small Maf proteins (sMaf). The Nrf2–sMaf heterodimer coordinately induces multiple cellular defense enzymes (12–14). This stress-mediated gene regulation system is referred to as the Keap1-Nrf2 system, which acts as a key regulator for protective responses against RS (6,7).

Loss of Nrf2 activity increases urine nitrite and nitrate levels in streptozotocin (STZ)-induced diabetic mice (15), suggesting critical contributions of Nrf2 to diabetes complications. In addition, we found that Nrf2 prevents onset of diabetes mellitus by effects including multiple tissues, such as liver, skeletal muscle, and pancreatic β-cells (16). While we have demonstrated that Nrf2 preserves pancreatic islets of db/db mice, the mechanism for how Nrf2 protects β-cells has not been clarified. Therefore, in this study, we have focused on the examination of whether Nrf2 contributes to reduction of RS in β-cells.

While several chemical Nrf2 inducers have been reported to ameliorate STZ-induced β-cell damage (17,18), off-target effects of Nrf2 inducers have hampered further analyses of actual contributions of the Keap1-Nrf2 system to β-cell protection (19–21). In addition, as STZ probably does not primarily exert broad-spectrum effects in β-cell damage by RS (22), the STZ model gives rise to technical difficulties in assessing the RS involvement. Therefore, to demonstrate the roles of Nrf2 in β-cells, we used four genetically engineered mouse models in this study: 1) β-cell–specific Keap1-conditional knockout mice as a genetic Nrf2-induction model (23), 2) β-cell–specific Nos2 transgenic mice as an RS-mediated damage model (24), 3) conventional Nrf2-heterozygous knockout mice as a systemic Nrf2-reduction model (1), and 4) β-cell–specific Nrf2-conditional knockout mice (25) as an Nrf2-depletion model.

Exploiting these four mouse lines, we show here that Nrf2 markedly protects pancreatic β-cells from various RS. Genetic Nrf2-induction nicely reduced β-cell damage in RS-mediated damage model mice. In contrast, genetic Nrf2 depletion markedly enhanced β-cell damage. Our results provide direct lines of evidence that the Keap1-Nrf2 system is the crucial defense pathway for the physiological and pathological protection of pancreatic β-cells.

Keap1-conditional knockout mice were previously described (23). The rat Ins2-Cre (RIP-Cre) mouse line was generated independently by a transgene harboring rat Ins2 promoter (BamHI-XmaI fragment; –695/+22), Cre cDNA, SV40 splice donor/acceptor sites, and SV40 polyA signal (26). Transgenic mice expressing chimeric DNA combining rat Ins2 promoter and Nos2 gene cDNA (iNOS-Tg mice) were previously described (24). Two iNOS-Tg mice lines, which showed high (line H) and low (line L) levels of inducible nitric oxide synthase (iNOS) expression, were used in this study. The Nrf2-conventional knockout (or Nrf2-LacZ-knockin) mouse line, which was knocked-in LacZ reporter into the Nrf2 locus, expressed partial Nrf2 and β-galactosidase fusion protein (Nrf2-β-gal) similar to the pattern of expression of wild-type Nrf2 (1,27). Nrf2-conditional knockout mice were previously described (25). A loxP-flanked stop-cassette reporter and Rosa26-tdTomato and Alb-Cre mice were supplied from The Jackson Laboratory (28,29). Mouse crossings and background information are described in Supplementary Table 1. Recombination of Keap1^flox/flox allele was determined by PCR methods using specific primers (5′-GACTAAAGCAGGAGATCGC-3′, 5′-GTTTGAGGCCAGTTTGGTC-3′, and 5′-ACAGCTCCTCGCCCTTGCTC-3′).

Blood glucose levels were determined using OneTouch UltraView (LifeScan). Plasma insulin and glucagon levels were determined using Insulin ELISA Kit (Morinaga Institute) and Glucagon EIA Kit (Yanaihara Institute) according to the manufacturers’ instructions. Glucose tolerance test (GTT) and insulin tolerance test (ITT) were previously described (16). Homeostasis model assessment as an index of insulin resistance (HOMA-IR) was calculated as plasma insulin times blood glucose per 405.

Immunohistochemistry was performed using monoclonal anti-insulin antibody (K36AC10, Sigma-Aldrich; 1:1,000 dilution) and polyclonal anti-NQO1 antibody (Abcam; 1:400 dilution) as previously described (16). For fluorescence staining, sections were stained with polyclonal anti-insulin (Abcam; 1:200 dilution) and anti-glucagon (DAKO; 1:100 dilution), anti-Ki67 (DAKO; 1:50 dilution), and anti-β-galactosidase (Millipore; 1:500 dilution) antibodies. For 8-hydroxy-2'-deoxyguanosine (8-OHdG) staining, pancreata were fixed in Bouin’s solution and immunostained with mouse monoclonal anti-8-OHdG antibody (N45.1, Nikkenzail; 1:20 dilution). Insulin-positive areas of sections and islet size were measured using ImageJ software (30).

Isolation of islets was previously described (16). A batch of 10 isolated islets was pretreated with RPMI-1640 containing 2.8 mmol/L glucose for 30 min. Islets were incubated in RPMI-1640 containing 16.7 mmol/L glucose or 60 mmol/L KCl for 1 h; supernatants were used to measure insulin secretion.

RNA extraction and quantitative PCR, previously described (16), were performed using primers in Supplementary Table 1. Immunoblot analysis was performed with anti-3-Nitrotyrosine (39B6, Abcam; 1:1,000 dilution) and anti-α-tubulin (Sigma-Aldrich; 1:12,000 dilution) as previously described (16,31,32).

RINm5F cells (33) were maintained at 37°C in a humidified incubator with atmosphere of 5% CO₂ and cultured in RPMI-1640 containing 11.1 mmol/L glucose, 10% FBS, 100 units/mL penicillin, and 100 µg/mL streptomycin. MIN6 cells were kindly provided by Jun-ichi Miyazaki (Osaka University) and cultured in Dulbecco’s modified Eagle’s medium containing 25 mmol/L glucose, 13% FBS, 100 units/mL penicillin, 100 µg/mL streptomycin, and 100 µmol/L 2-mercaptoethanol.

After treatment of islets and RINm5F cells with oleanolic acid 1-[2-cyano-3, 12-dioxooleana-1,9(11)-dien-28-oyl]-imidazole (CDDO-Im; Mochida Pharmaceuticals), islets and cells were treated with NOR3 (Dojindo) or SIN-1 (Dojindo) and then loaded with 2',7'-dichlorodihydrofluorescin-diacetate (DCF-DA) (Sigma-Aldrich) or 2-[6-(4'-hydroxy)phenoxy-3H-xanthene-3-on-9-yl]benzoic acid (HPF) (Sekisui-Medical) (34). To analyze islets, fluorescence images were obtained using a laser-scanning stereoscopic-microscope (Keyence). For analyses with culture cells, fluorescent intensities were analyzed using FACSCalibur flow-cytometer (BD Biosciences). In cytokine-mediated iNOS-induction model, isolated islets were treated with cytokines (35) (tumor necrosis factor-α [Wako Pure Chemicals], interferon-γ [Wako Pure Chemicals], and interleukin-1β [BD Biosciences]) and nitric oxide (NO) synthase inhibitors (Nω-nitro-L-arginine methyl-ester-hydrochloride [Sigma-Aldrich], N^G-methyl-L-arginine acetate [Calbiochem], 1400 W [Sigma-Aldrich], and aminoguanidine [Wako Pure Chemicals]).

A batch of 50 isolated islets was incubated in 300 μL RPMI-1640/10% FBS without phenol red (Sigma-Aldrich). Media (50 μL) were collected at indicating times. Nitrite levels were analyzed using Nitric Oxide Assay Kit (Thermo Fisher) following the manufacturers’ instructions.

RINm5F cells were incubated with NOR3 (NO donor), N-acetyl cysteine (NAC; Sigma-Aldrich), and glutathione (GSH; Wako Pure Chemicals). After staining cells with 0.4% trypan blue, cells were counted by Countess automated cell counter (Life Technologies). Apoptosis was evaluated by TUNEL staining using In Situ Cell-Death Detection Kit (Roche Diagnostics), according to the manufacturer’s instructions.

All data are presented as the mean ± SEM. Statistical analyses were performed using Student’s t test or ANOVA followed by Fisher’s least significant difference post hoc test for multiple comparison.

To assess the cytoprotective effects of Nrf2 in β-cells, we generated β-cell–specific Keap1-conditional knockout (Keap1-βCKO [RIP-Cre::Keap1^flox/flox]) mice. We also used iNOS-Tg mice as a β-cell–specific RS-mediated damage model (24). The iNOS-Tg (line H) mice showed severe β-cell damage (Supplementary Fig. 1A). The β-cell damage in these mice actually caused them to develop elevated glucose levels (Supplementary Fig. 1B).

We next examined the pancreas of the iNOS-Tg (line H) and Keap1-βCKO double mutant (iNOS-Tg::Keap1-βCKO) mice. Of note, 3,3′-diaminobenzidine-stained insulin-positive area and islet size were restored in iNOS-Tg::Keap1-βCKO mice, and both were comparable compared with those in wild-type mice (Fig. 1A–C). In contrast, insulin-positive area and islet size in iNOS-Tg::RIP-Cre::Keap1^flox/+ (iNOS-Tg::Keap1-NegC) was markedly reduced compared with wild-type mice (Fig. 1A–C). We have used iNOS-Tg::RIP-Cre::Keap1^flox/+ mice as negative control mice (iNOS-Tg::Keap1-NegC) throughout this study, but we also used iNOS-Tg::RIP-Cre mice for this purpose additionally where specified. To verify the Nrf2 induction in Keap1-deficient mice, we attempted immunohistochemical detection of Nrf2 protein. However, we could not get stable signals from anti-Nrf2 antibody. Therefore, we conducted NQO1 immunohistochemistry, as NQO1 was one of the faithful target genes of Nrf2. We found that NQO1 was strongly expressed in iNOS-Tg::Keap1-βCKO islets but not in iNOS-Tg::Keap1-NegC islets (Fig. 1D), strongly supporting our contention that Nrf2 was induced by Keap1-conditional knockout.

We also evaluated oxidative DNA-adduct formation in iNOS-Tg(line-L)::Keap1-βCKO mice by 8-OHdG immunohistochemistry. In iNOS-Tg::Keap1-NegC mice, 8-OHdG-positive β-cells were frequently observed, but these cells were rarely observed in iNOS-Tg::Keap1-βCKO mice (Fig. 1E). These results indicate that Nrf2 protects β-cells against oxidative DNA-adduct formation.

We next evaluated pancreatic sections fluorescent double stained with insulin and glucagon antibodies. Fluorescent insulin-positive area (Fig. 2A, green fluorescence) was markedly decreased in iNOS-Tg and iNOS-Tg::Keap1-NegC mice, but the insulin-positive area strongly recovered in iNOS-Tg::Keap1-βCKO islets (Fig. 2B). Glucagon-positive area (Fig. 2A, red fluorescence) in iNOS-Tg::Keap1-NegC islets was slightly increased, and that in iNOS-Tg::Keap1-βCKO islets was similar to iNOS-Tg::Keap1-NegC islets (Fig. 2C). We also examined glucagon-positive cell arrangement in islets. In wild-type pancreatic sections, glucagon-positive cells were localized in the periphery of islets (Fig. 2A, top-middle panel). In iNOS-Tg and iNOS-Tg::Keap1-NegC mouse islets, a condensed arrangement pattern of glucagon-positive cells in the center of the islets was observed. Intriguingly and against our expectation, arrangement pattern of glucagon-positive cells did not recover in iNOS-Tg::Keap1-βCKO mice to the wild-type peripheral positioning pattern (Fig. 2A, bottom-middle panel). These results thus unequivocally demonstrate that Nrf2 protects insulin-positive cells from RS-mediated damage.

We next wished to delineate whether Nrf2 supports β-cell functional restoration. To this end, we examined insulin secretion from isolated islets. Despite comparable basal insulin release, both glucose- and potassium-stimulated insulin secretions from iNOS-Tg::Keap1-βCKO islets were markedly preserved (Fig. 3A). Insulin contents of iNOS-Tg::Keap1-βCKO islets increased approximately two-fold compared with iNOS-Tg::Keap1-NegC islets (Fig. 3B). These data indicate that Nrf2 provided significant protection to damaged β-cells.

We evaluated insulin-related gene expression in isolated islets. We found that Ins1, Ins2, and Kir6.2 gene levels were upregulated in iNOS-Tg::Keap1-βCKO islets (Fig. 3C); in contrast, SUR1 was decreased. Although the reason for SUR1 repression was not clear, these results support our contention that Nrf2 signaling activation indeed preserves β-cell function in iNOS-Tg mice.

We next evaluated the expression profiles of several antioxidant-enzyme genes in Keap1-βCKO islets. Nqo1, Gstp1, Gpx2, and Txnrd1 mRNA levels were upregulated in Keap1-βCKO mice (Fig. 4A). Importantly, the induction of these genes was reduced partially and completely by the crossing of Keap1-βCKO mice with Nrf2^+/− and Nrf2^−/− mice, respectively (Fig. 4A). We examined the NQO1 protein levels in Keap1-βCKO islets for monitoring Nrf2 induction (Fig. 4B).

To examine how NO influences the Keap1–Nrf2 system in β-cells, we next examined antioxidant-enzyme gene expression in isolated islets in the presence of NO donors. We used two NO donors: SIN-1 and NOR3 (36). Both SIN-1 and NOR3 highly induced Nqo1, HO-1, and Txnrd1 expression levels in islets (Fig. 4C). We also used an electrophile, diethyl maleate (DEM), as a positive control. We found that DEM induced all six antioxidant-enzyme genes (Fig. 4C). To ascertain the Nrf2 contribution in these gene inductions, we examined islets from Nrf2-knockout mice. The upregulation of Nqo1 by SIN-1 and DEM were both clearly canceled by Nrf2 depletion (Fig. 4D). These results unequivocally demonstrate that Nrf2 induces antioxidant-enzyme genes in islets and that NO contributes to Nrf2 induction.

We next determined RS levels in iNOS-Tg mouse islets using DCF-DA. DCF-fluorescent intensity was strongly elevated in iNOS-Tg islets (Supplementary Fig. 2A and B). To ascertain whether iNOS-derived RS actually increases DCF fluorescence, we also measured the DCF fluorescence of three cytokine-treated islets (tumor necrosis factor-α/interferon-γ/interleukin-1β). We found that cytokines increased Nos2 expression (data not shown) and DCF fluorescence (Supplementary Fig. 2C) while iNOS inhibitors suppressed the cytokine-induced DCF fluorescence (Supplementary Fig. 2D). These data nicely confirm the notion that RS levels indeed increase in iNOS-Tg islets.

We next examined the role that Nrf2 plays in the repression of RS in islets. We first determined RS in isolated islets using DCF-DA and found that RS were similar in CDDO-Im–treated and vehicle-treated groups (Fig. 5A). However, in the presence of SIN-1, the increase in RS was markedly suppressed by CDDO-Im (Fig. 5A). Importantly, the suppression of RS by CDDO-Im was completely abrogated in Nrf2-knockout mice (Fig. 5B). To increase quantitative accuracy, we additionally examined DCF fluorescent intensities in RINm5F and MIN6 cells by flow cytometry. The analyses clearly revealed that CDDO-Im reduced SIN-1–induced RS in both RIN5mF (Fig. 5C) and MIN6 (Supplementary Fig. 3A) cells.

We next worked to determine which RS was decreased by Nrf2. Since HPF is more specific to hydroxyl radical and less sensitive to peroxynitrite than DCF-DA (34), we used HPF in RIN5mF cells. Although both SIN-1 and NOR3 weakly increased HPF fluorescence, DCF-DA showed much higher fluorescent intensity than HPF (Supplementary Fig. 3B). We found that the mean fluorescent intensity (MFI) of HPF was mildly augmented by SIN-1 (Supplementary Fig. 3C). The SIN-1–induced HPF fluorescence was suppressed by CDDO-Im. To determine peroxynitrite levels, we examined immunoblot analysis for nitrotyrosine. We found a 70 kDa nitrotyrosine-bound protein in the SIN-1–treated RINm5F cells, and CDDO-Im markedly decreased the nitrotyrosine levels (Supplementary Fig. 3D). To validate Nrf2 induction, we next evaluated Nqo1 mRNA levels and verified that CDDO-Im strongly increased Nqo1 expression in RINm5F and MIN6 cells (Supplementary Fig. 3E and F) as well as isolated islets as previously reported (16). These data indicate that Nrf2 represses mainly peroxynitrite and partially hydroxyl radical.

We then examined NO production in iNOS-Tg (line L) mouse islets. To this end, we measured the nitrite release from islets into the culture media. Nitrite release levels from isolated islets did not significantly differ in iNOS-Tg::Keap1-βCKO islets compared with iNOS-Tg::Keap1-NegC islets (Supplementary Fig. 4). These results support the hypothesis that Nrf2 acts to decrease RS levels independently of NO production. The data also suggest that Nrf2 does not suppress NO production in iNOS-Tg islets.

To evaluate the effect of β-cell protection mediated by Nrf2 on whole-body metabolism, we next evaluated glucose levels of iNOS-Tg::Keap1-βCKO mice. We found that glucose levels of iNOS-Tg::Keap1-βCKO mice were lower than those of iNOS-Tg::Keap1-NegC mice (Fig. 6A), but we could not find a significant difference in insulin levels of 10-week-old mice between the two groups (Fig. 6B). Therefore we examined glucose metabolism until 47 weeks of age. Notably, insulin levels were significantly increased in 20-week-old iNOS-Tg::Keap1-βCKO mice compared with iNOS-Tg::Keap1-NegC mice (Fig. 6B) without changing glucagon levels (Supplementary Fig. 5A). In contrast, insulin levels were decreased in 47-week-old iNOS-Tg::Keap1-βCKO mice, with improvement of HOMA-IR (Supplementary Fig. 5B), indicating the enhancement of insulin sensitivity. iNOS-Tg::Keap1-βCKO mice showed lower body weight than iNOS-Tg::Keap1-NegC mice without changing food intake (Supplementary Fig. 5C and D). To strictly evaluate glucose metabolism of iNOS-Tg::Keap1-βCKO mice, we next performed GTTs and ITTs. We found that glucose levels and areas under the curve (AUCs) of iNOS-Tg::Keap1-βCKO mice in GTT and ITT were decreased (Fig. 6C and D). Thus these data support our contention that iNOS-Tg::Keap1-βCKO mice show lowered glucose levels due to the enhanced insulin sensitivity with the increase in insulin levels.

It has been reported that RIP-Cre may cause abnormalities in glucose regulation (37–39), but the RIP-Cre mice we used in this study (26) were a different transgenic line independently established from the previously reported lines. To ascertain tissue specificity of our RIP-Cre in mice, we evaluated recombination of Keap1^flox/flox and Rosa26-tdTomato. We found both the recombination of Keap1^flox/flox alleles in islets by using PCR methods and repression of Keap1 mRNA levels (Supplementary Fig. 6A and B). The recombination was also observed in the brain, but not in the liver, skeletal muscle, adipose tissue, heart, or spleen. As another control, we used Alb-Cre::Keap1^flox/flox mice and observed sharp liver specificity. Consistently, the RIP-Cre line we used in this study also directed recombination of Rosa26-tdTomato (28) in pancreas and hypothalamus (Supplementary Fig. 6C and D). Since hypothalamic Cre expression may affect metabolism (36,37), we also performed GTT and ITT. Blood glucose levels in GTT and ITT were comparable in RIP-Cre and wild-type mice (Supplementary Fig. 6E and F). These results indicate that the RIP-Cre transgene we used per se did not affect glucose metabolism or insulin secretion in these mice.

Thus far, we have shown that Nrf2 gain-of-function mutation gives rise to β-cell protection from RS. To clarify how Nrf2 loss-of-function mutation affects β-cell function of iNOS-Tg mice, we next examined the progeny of the iNOS-Tg (line L) mice crossed to Nrf2^+/− mice.

We observed small islets in the pancreas of iNOS-Tg::Nrf2^+/+ mice concomitant with regular-sized islets (Fig. 7A). In contrast, we did not observe small islets in wild-type (Nrf2^+/+) mouse pancreas. To our surprise, in iNOS-Tg::Nrf2^+/− mice, we found that regular-sized islets were markedly decreased, and small islets became predominant in iNOS-Tg::Nrf2^+/− mutant mice (Fig. 7A). The NQO1 repression was confirmed in iNOS-Tg::Nrf2^+/− islets (Supplementary Fig. 7A). We executed quantitative analyses of islet size in insulin-positive area. We classified islets into four groups according to their size as described in Fig. 7B. In wild-type pancreas, group I large islets were frequently observed (Fig. 7B). In contrast, group I islets were decreased in iNOS-Tg::Nrf2^+/+ pancreas, and numerous group IV islets emerged in iNOS-Tg::Nrf2^+/− pancreas. These results indicate that islet size becomes dramatically smaller in the absence of Nrf2. We next performed GTT using iNOS-Tg::Nrf2^+/− mice. Blood glucose levels in GTT were comparable in Nrf2^+/− and Nrf2^+/+ mice, but iNOS-Tg::Nrf2^+/+ mice showed moderately impaired glucose tolerance. Importantly, blood glucose levels were significantly aggravated in iNOS-Tg::Nrf2^+/− mice (Fig. 7C). These results further support the contention that Nrf2 contributes to preservation of β-cell function against RS.

It has been reported that Nrf2-null mice exhibit lower blood glucose levels in GTT compared with wild-type mice (16,40,41). These results argue that Nrf2 knockout and general Nrf2 loss of function may cause systematic perturbations of glucose regulation. Therefore, to exclude effects of systemic Nrf2 deletion on glucose regulation, we generated β-cell–specific Nrf2-knockout mice (Nrf2-βCKO (RIP-Cre::Nrf2^flox/flox)) crossed with iNOS-Tg (line L) mice.

To this end, we switched mouse background from ICR to C57BL/6J, because Nrf2-conditional knockout mice were in C57BL/6J background. We also changed the background of both RIP-Cre and iNOS-Tg mice to C57BL/6J. During this background switching, we unexpectedly found that the number of large islets increased in pancreas of iNOS-Tg::Nrf2^flox/flox mice in C57BL/6J background compared with Nrf2^flox/flox mice (Fig. 7D). These findings are in clear contrast to observations in ICR background iNOS-Tg::Nrf2^+/+ mice (Fig. 7A), and the reason for this difference is presently unclear. However, an important observation was that the iNOS-Tg::Nrf2-βCKO pancreas displayed small islets much more abundantly than iNOS-Tg::Nrf2^flox/flox pancreas (Fig. 7D) with repression of NQO1 expression (Supplementary Fig. 7B). We evaluated size of islets using insulin-positive staining. Group I islets were more frequently detected in iNOS-Tg::Nrf2^flox/flox mice than in Nrf2^flox/flox mice (Fig. 7E), and group I islets dramatically decreased and group III islets increased in iNOS-Tg::Nrf2-βCKO mice (Fig. 7E). These results indicate that RS severely affects β-cells in Nrf2-deficient conditions.

We also examined blood glucose levels of iNOS-Tg::Nrf2-βCKO mice using the intraperitoneal glucose tolerance test (ipGTT) and found that the iNOS-Tg::Nrf2-βCKO mice showed impaired glucose tolerance (Fig. 7F). Plasma insulin levels in iNOS-Tg::Nrf2-βCKO mice were also decreased compared with those in Nrf2^flox/flox and iNOS-Tg::Nrf2^flox/flox mice (data not shown). Thus supporting results using conventional Nrf2-heterozygous knockout mice and β-cell–specific Nrf2-conditional knockout mice display small islets and impaired glucose tolerance in response to RS. These results strongly argue that Nrf2 contributes to the preservation of β-cells.

We explored mechanisms for how Nrf2 protects β-cells. Nrf2 regulates GSH-related genes in islets (16), and CDDO-Im indeed increased Gclc, Gsta4, Gstm1, and Gpx2 expression in MIN6 cells (Fig. 8A). To demonstrate importance of GSH in β-cell protection, we examined roles of GSH on cell death and apoptosis utilizing trypan blue and TUNEL staining. Although 2-mercaptoethanol was required to maintain MIN6 cells, it was also reported to affect GSH levels in β-cells (42). Therefore, we changed β-cell line to RINm5F cells. NOR3 increased trypan blue-positive RINm5F cells, which was dramatically suppressed by NAC and GSH in a dose-dependent manner (Fig. 8B). In addition, NOR3 dramatically increased TUNEL-positive RINm5F cells (Fig. 8C). Of note, NAC and GSH strongly decreased the TUNEL-positive cells. These data indicate that one of the mechanisms that protect β-cells is GSH-mediated repression of apoptosis.

To clarify how Nrf2 preserves islet size, we evaluated cell proliferation by Ki67. iNOS-Tg::Keap1-NegC islets rarely had Ki67-positive cells. The percentage of Ki67-positive islets was increased in iNOS-Tg::Keap1-βCKO (Supplementary Fig. 8). These data indicated that one of the reasons that Nrf2 preserves islet size is the β-cell proliferation.

Although CDDO-Im markedly represses RS in vitro, it has not been confirmed whether oral administration of Nrf2 inducers activates the Keap1-Nrf2 system in β-cells in vivo. Therefore, we finally evaluated Nrf2 induction by the Nrf2 nuclear accumulation in Nrf2-LacZ-knockin mice (1,27). In the absence of CDDO-Im, Nrf2–β-gal fusion protein was localized in cytoplasm of Nrf2-LacZ-knockin β-cells (Supplementary Fig. 9, left-hand panels). Importantly, oral administration of CDDO-Im accumulated Nrf2–β-gal in the nucleus (right-hand panels). These results unequivocally demonstrate that the oral administration of Nrf2-inducer activates the Nrf2 signal.

To address whether and how the Keap1-Nrf2 system contributes to protection of pancreatic β-cells (16), in this study we used four genetic model mouse lines and have clearly demonstrated that Nrf2 protects β-cells from RS-mediated damage. We have used iNOS-Tg mice as a model mouse line to assess RS-mediated damages in β-cells. We have decided to use this line of mice since RS directly damaged β-cells in this line of mice (24). Of note, we found that genetic Nrf2 induction strongly preserves the insulin-positive area of the pancreas in iNOS-Tg mice. We also found that Nrf2 prevents deleterious effects of RS in the islets of iNOS-Tg mice and that Nrf2 reduces RS levels in NO-donor–treated RINm5F cells. Our present results provide compelling lines of evidence that the Keap1-Nrf2 system contributes to preservation of systemic glucose homeostasis through protecting β-cells.

DCF fluorescence is induced by reactive oxygen species (ROS; e.g., hydroxyl radical [HO•], hypochlorite anion [^–OCl], and hydrogen peroxide [H₂O₂]) and reactive nitrogen species (e.g., peroxynitrite [ONOO^–]) (34,43). In this study, we found that Nrf2 strongly decreases DCF fluorescence. Importantly, Nrf2 was also found to mildly attenuate HPF fluorescence. It has been reported that DCF-DA sensitively detects reactive nitrogen species peroxynitrite and ROS hydroxyl radical, while, in contrast, HPF detects more specifically ROS hydroxyl radical than DCF-DA (34). Therefore Nrf2 seems to suppress mainly the peroxynitrite level. We also have shown that Nrf2 protects β-cells from apoptosis by the induction of GSH-related genes. It has been reported that peroxynitrite indeed induces apoptosis (44), but peroxynitrite is scavenged by GSH (45), a major target of Nrf2-mediated gene regulation. Based on these observations, we surmise that peroxynitrite is a major RS that is scavenged in Nrf2-mediated β-cell protection.

In this study, iNOS-Tg::Keap1-βCKO mice showed complicated regulations over glucose and insulin levels. Most significantly, the increase of Nrf2 in iNOS-Tg::Keap1-βCKO mice affects differentially the glucose and insulin levels by aging. We envisage two plausible elements that cause this complexity. One is the decreasing changes of β-cell damage in iNOS-Tg mice with aging, and the other is the incremental changes in insulin sensitivity of iNOS-Tg::Keap1-βCKO mice with aging. Available lines of evidence suggest that these two factors may lead to the dissociated relationship between glucose levels and insulin levels and also cause the large deviations in individual iNOS-Tg mice.

In this study, Ins1, Ins2, and Kir6.2 levels are found to be increased in iNOS-Tg::Keap1-βCKO islets, but in contrast, SUR1 levels are decreased. This observation is intriguing, as Nrf2 is generally accepted to activate transcription of genes (6,7) and other cap’n’collar family factors, such as Bach1 (46) or p45 (47), act to repress Nrf2 activity. However, we have experienced that for some genes, Nrf2 works to repress their expression, such as gluconeogenesis-related genes (16). The precise mechanisms for the Nrf2-mediated activation and/or repression of gene transcription remain to be clarified.

To clarify how Nrf2 preserves islet size in iNOS-Tg mice, we have examined β-cell proliferation and apoptosis. One of the salient findings in this study is that in β-cells of iNOS-Tg::Keap1-βCKO islets, Ki67-positive nuclei are increased, and concomitantly, β-cells are protected from apoptosis. Indeed, Nrf2 function in cell proliferation was recently identified (48). Based on these observations, we conclude that Nrf2 preserves islet size by both enhancement of β-cell proliferation and repression of β-cell apoptosis in iNOS-Tg mice. In conclusion, this study has revealed that the Keap1-Nrf2 system is a key regulator in the protection of pancreatic β-cells. We also recognize that the Keap1-Nrf2 system serves as a promising target for the prevention of pancreatic β-cell damage.

Acknowledgments. The authors thank Hiroshi Okamoto (Tohoku University) for the iNOS-Tg mice, Tkaaki Akaike (Tohoku University) for advice in NO studies, Jun-ichi Miyazaki (Osaka University) for providing MIN6 cells, Sayoi Inomata (Tohoku University) for assisting with mouse genotyping, Eriko Naganuma (Tohoku University) and Fumiko Date (Tohoku University) for assisting with immunohistochemistry, and the Biomedical Research Core of Tohoku University Graduate School of Medicine for technical support.

Funding. This work was supported in part by the Japanese Foundation for Applied Enzymology (A.U.); Grants-in-Aid for Scientific Research on Innovative Areas and Scientific Research from the Ministry of Education, Science, Sports and Culture (M.Y.); Grants-in-Aid for Scientific Research from the Japan Society for the Promotion of Science (grant numbers 24249015 [M.Y.], 90396474 [A.U.]); the Core Research for Evolutional Science and Technology from the JST (M.Y.); the Takeda Foundation (M.Y.); and the Naito Foundation (M.Y.).

Duality of Interest. M.Y. has received grant support from Mochida Pharmaceutical Co. Ltd. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. Y.Y. contributed to research data and preparation of the manuscript. T.F. contributed to research data and the discussion. A.S. contributed to the preparation of the iNOS-Tg mouse lines and research using MIN6 cells. H.K. and T.T. contributed to the discussion. J.P. contributed to the preparation of the Nrf2-conditional knockout mouse line. A.U. contributed to study design, research data, and preparation of the manuscript. M.Y. contributed to study design and preparation of the manuscript. A.U. and M.Y. are the guarantors of this work and, as such, had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.