Nuclear Factor-κB Activity in β-Cells Is Required for Glucose-Stimulated Insulin Secretion

IκBαM cDNA (Clontech, Palo Alto, CA) was cloned after the Ipf1/Pdxl promoter (9) and followed by a SV40 polyA site. Transgenic mice were generated by pronuclear injection of the purified fragment (1.8 ng/ml) into F2 hybrid oocytes from B6/CBA parents (M&B) (10). The genotype was determined by PCR analysis of genomic DNA extracted from tail biopsies using primers insulin promoter factor (IPF) 15′-3(GGGAAGAGGAGATGTAGACTT), IKBAM-R(GGTCAGTGCCTTTTCTTC), and IPF1-AR(GAGCTGAGCTGGAAGGT). Ipf1/IκBαM integration was analyzed by genomic Southern blot using a probe complementary to the Ipf1 promoter. Four independent transgenic lines showed blunted GSIS, and two of these were selected for further analysis.

MIN6 cells were transfected with pNF-Luc (Promega), pMT21/lacZ, Ipf1/IκBαM, or empty vector using Lipofectamin according to the manufacturers recommendations. The cells were incubated overnight in Dulbecco’s modified Eagle’s medium, 5.5 mmol/l glucose, and 10% fetal bovine serum (FBS). The cells were then stimulated for 2 h with pyrollidine dithiocarbamate (PDTC; 100 μmol/l), and luciferase activity was measured using a luciferase assay system (Promega). All values were normalized to β-galactosidase activity from cotransfected pMT21/lacZ. Islets were isolated and recovered overnight in RPMI 1640 with 10% FBS (11) before experiments. For insulin secretion measurement, islets were equilibrated 1 h in 1 × HEPES medium (3.3 mmol/l glucose) and transferred to 1 × HEPES medium containing 3.3 or 20 mmol/l glucose. PDTC treatment was started during ON recovery and continued throughout the experiment. The supernatant and islets were analyzed for insulin content using enzyme-linked immunosorbent assay (see below). For cytokine stimulation, islets were stimulated for 8 h with interleukin-1β (1 ng/ml) and γ-interferon (1 μg/ml) in RPMI 1640 with 10% FBS.

Immunohistochemistry was performed essentially as described (12) on neonatal and adult (aged 10–12 weeks) B6/CBA mice pancreas. Primary antibodies used were rabbit anti-IκBα (sc-847; Santa Cruz Biotechnology), rabbit anti-IPF1/pancreatic duodenal homeobox 1 (PDX1) (13), guinea pig anti-insulin (Linco), guinea pig anti-glucagon (Linco), rabbit anti-glucagon (Linco), rat anti-somatostatin (Bender MedSystems), rabbit anti-GLUT2 (AB1342; Chemicon), rabbit anti–sulfonylurea receptor 1 (SUR1) (14), and rabbit anti-Kir6.2 (15). The following secondary antibodies were used: Cy3-conjugated goat anti-rabbit IgG, dichlorotriazine fluorescein–conjugated goat anti–guinea pig IgG, and Cy3-conjugated streptavidin (The Jackson Laboratory). Apoptosis was detected by terminal deoxynucleotidyl transferase–mediated dUTP nick-end labeling (TUNEL) assay (Roche Diagnostics) or by staining for the apoptotic marker–cleaved caspase 3 (Cell Signaling Technology).

Neonatal pancreas from wild-type (n = 4) and NIβ (n = 3) mice were sectioned. Every 10th section was selected for analysis. The different endocrine cell types were detected with antibodies against insulin, glucagon, somatostatin, and pancreatic polypeptide. Isl1 antibodies (16) were used to label all endocrine cells. The ratio between the total area of immunoreactivity and the total area of 4′,6 diamidino-2-phenylindole staining was calculated by image analysis software (Image Pro Plus; Media Cybernetics) and used to evaluate the abundance of the different endocrine cell types.

Overnight-fasted (GSIS test and arginine test) or nonfasted mice were injected intraperitoneally with the following substances (dose/kg body wt): glucose (1 g/kg), recombinant human insulin (0.75 units/kg; Novo Nordisk), glybenclamide (5 mg/kg G-0639; Sigma), carbachol (0.16 μmol/kg C-4382; Sigma), and l-arginine (1 g/kg A-5131; Sigma). Blood samples were obtained from the tail vein. Blood glucose levels were measured using a Glucometer Elite (Bayer). Total pancreatic insulin was extracted using acid ethanol (75% ethanol, 0.2 mol/l HCl) and measured using a sensitive rat insulin radioimmunoassay kit (Linco). Total pancreatic protein concentration was determined using a protein assay (Bio-Rad). Serum insulin levels were measured using enzyme-linked immunosorbent assay (Crystal Chem). Statistical significance was calculated using Student’s t test.

Cytoplasmic protein extract was prepared from isolated islets and analyzed using Western blot. Endogenous IκBα and mutant IκBαM were detected using anti-IκBα (sc-847; Santa Cruz Biotechnology) and Supersignal West Pico chemiluminescent substrate (Pierce) and quantified using Imagequant software. NIβ lines that expressed high levels of IκBαM were chosen for further analysis. cDNA was prepared from total RNA prepared from freshly isolated islets (Fig. 5) or cultivated islets (Fig. 2) essentially as described (9). Real-time PCR analysis was performed using an ABI Prism 7000 sequence detection system and SYBR Green PCR Master Mix (ABI) according to the manufacturer’s recommendations. Expression of the 60S acidic ribosomal protein P0 (Arbp) (Figs. 2 and 5) and β-2-microglobulin (b2M) (Fig. 5) was used to normalize expression levels.

A commonly used, specific, and effective strategy to suppress NF-κB activation is to express a dominant active mutant of IκBα, IκBαM, which acts by sequestering NF-κB dimers in an inactive cytoplasmatic complex (17,18). In addition, the antioxidant PDTC has been widely used to inhibit NF-κB activity in cells (19), although the mechanism of action of this agent remains poorly understood. NF-κB activity in MIN6 insulinoma cells was monitored by transfection of a luciferase reporter driven by five tandem NF-κB consensus binding sites. The expression of the NF-κB reporter construct was efficiently reduced in MIN6 cells when cotransfected with an IκBαM construct, driven by the Ipf1 promoter (Ipf1/IκBαM) (Fig. 1A). Similarly, the chemical NF-κB inhibitor PDTC also impaired NF-κB transcriptional activity in MIN6 cells (Fig. 1A). Thus, both experimental approaches inhibited NF-κB activity in MIN6 insulinoma cells.

To test whether attenuation of NF-κB activity affects GSIS, we monitored the release of insulin in isolated islets treated with PDTC. Islets elicited an almost fivefold increase in secreted insulin when exposed to high glucose. In contrast, high glucose failed to stimulate insulin secretion from islets exposed to PDTC (Fig. 1B), suggesting that NF-κB activity may be required in isolated islet cells for efficient GSIS.

Although PDTC is a known and widely used inhibitor of NF-κB, we cannot exclude that it also may affect other, NF-κB–unrelated targets. Thus, to selectively inhibit NF-κB and elucidate whether NF-κB activity is required for β-cell function and insulin secretion in vivo, we next generated transgenic mice expressing the dominant-active mutant of IκBα, IκBαM, also known as the NF-κB super-repressor, under the control of the Ipf1 promoter. The resulting Ipf1/IκBαM transgenic mice are hereafter denoted NIβ (for NF-κB inhibition in β-cells). To confirm that IκBαM also suppresses NF-κB activation when expressed in the islets of the resulting NIβ mice, we analyzed the expression of cytokine-induced and NF-κB–dependent target genes in wild-type and transgenic islets after exposure to cytokines. Islets isolated from NIβ mice showed a 40–70% reduction of the known NF-κB–dependent target genes iNOS, MnSOD, MCP-1, and IκBα (20–23) after stimulation with interleukin-1β and γ-interferon (Fig. 2E). These data provide evidence that expression of IκBαM impairs NF-κB activity not only in insulinoma cells but also in β-cells of NIβ mice. A comparison of IκBαM expression in freshly isolated transgenic islets compared with in vitro–cultivated transgenic islets showed that the expression of IκBαM was reduced by 95% on in vitro cultivation (Fig. 2E), suggesting that the attenuation of NF-κB activity is more efficient in vivo.

NIβ mice were lean and fertile but showed elevated fasted and nonfasted blood glucose levels (i.e., mild diabetes) when compared with littermates (Table 1). To further examine the ability of the NIβ mice to control blood glucose levels, they were challenged with a pulse of exogenous glucose. Blood glucose in control mice typically peaked after 30 min and had returned to normal levels after 90 min. Mice derived from founder A, expressing high levels of IκBαM (Figs. 2A–D), failed to reduce blood glucose levels, and the blood glucose levels were still elevated after 180 min (Fig. 3). Mice derived from founder C, which expressed intermediate levels of IκBαM (Figs. 2C and D), also showed impaired glucose tolerance, although the blood glucose levels in these mice had declined by 180 min (Fig. 3). These results provide evidence that attenuation of NF-κB activation in β-cells in vivo leads to impaired glucose tolerance and mild diabetes.

To test whether the impaired glucose homeostasis in NIβ mice reflected perturbations in insulin secretion, we monitored insulin levels in NIβ mice at nonfasted conditions as well as serum insulin levels in response to glucose challenge. NF-κB activity is sensitive to oxidative, chemical, and physical stress (6), and in vitro cultivation of isolated transgenic islets also resulted in severely reduced expression of IκBαM (Fig. 2E). Thus, to avoid the stress associated with islet isolation and in vitro cultivation, and to ensure high-level expression of IκBαM, all analyses of insulin secretion were performed in vivo. Despite elevated glucose levels, serum insulin levels were significantly lower in nonfasted NIβ mice compared with control littermates (Table 1). Moreover, NIβ mice lacked both first- and second-phase insulin release in response to glucose challenge, whereas serum levels increased rapidly after glucose injection and peaked after 15 min in control mice (Figs. 4A and data not shown). Together, these results demonstrate that NIβ mice fail to secrete insulin in response to elevated blood glucose levels.

To exclude the possibility that the blunted GSIS observed in NIβ mice reflected a low number of β-cells, we determined β-cell number and total pancreatic insulin content. The NIβ mice exhibited a seemingly normally developed pancreas with islets (data not shown), and they showed a ∼25% reduction in the total number of endocrine cells as well as in total pancreatic insulin content when compared with control littermates (n = 7, P = 0.04). Conflicting reports suggest a role for NF-κB in both mediating and preventing cytokine-induced β-cell death (21–24). The observed 25% decrease in endocrine cell number in the NIβ mice could thus reflect an increase in apoptosis, which would support an antiapoptotic role for NF-κB in pancreatic β-cells. No increased cell apoptosis was, however, observed in neonatal or adult pancreata from NIβ mice when performing TUNEL assay or by staining for the apoptotic marker caspase 3 (data not shown), suggesting that the slight decrease in the number of endocrine cells in NIβ mice is not caused by increased apoptosis. Moreover, Bcl-2 mRNA was upregulated in isolated islets from NIβ mice, which would rather support a proapoptotic role of NF-κB in pancreatic β-cells (Fig. 5). The NIβ mice responded normally to exogenous insulin (Fig. 6), excluding both low numbers of β-cells and insulin resistance as major causes of the glucose intolerance and impaired GSIS. Collectively, these results provide evidence that attenuated NF-κB activation in β-cells severely perturbs GSIS but not β-cell generation.

Uptake of glucose into the β-cell, the first step in GSIS, is mediated by GLUT2, and decreased GLUT2 expression has been observed in a variety of diabetic animal models (25). Glut2 mutant mice display hyperglycemia, glucose intolerance, and a lack of first-phase insulin secretion in response to glucose (26). To assess whether perturbed GLUT2 expression contributed to the blunted GSIS of the NIβ mice, we compared GLUT2 expression in β-cells of the NIβ and wild-type mice at both the mRNA (Fig. 5) and protein (Fig. 7) level. Quantitative real-time PCR showed that the level of Glut2 expression was reduced by ∼80% in islets of adult NIβ compared with control mice (Fig. 5). Consequently, β-cells of both adult and neonatal NIβ mice failed to express high levels of GLUT2 protein (Fig. 7), providing evidence that suppression of NF-κB activation rather than hyperglycemia impairs GLUT2 expression in these mice. Thus, NF-κB activation is required for high-level GLUT2 expression in postnatal β-cells.

The transcription factors IPF1/PDX1 and hepatocyte nuclear factor-1α (HNF-1α) are also required for high-level expression of GLUT2 (27–30) in β-cells. To investigate whether the impaired Glut2 expression observed in the NIβ mice was the consequence of reduced expression of Ipf1/Pdx1 and/or HNF1-α, we monitored the expression of Ipf1/Pdx1, HNF1-α, and some of their proposed target genes (27–30). The expression of Ipf1/Pdx1 and HNF1-α and of the respective downstream genes fibroblast growth factor receptor 1c (FGFR1c) and prohormone convertase (PC1/3), as well as glucokinase, l-puryvate kinase (LPK), and mitochondrial 2-oxo-glutarate dehydrogenase (OGDH) appeared normal in the NIβ mice (Figs. 5 and data not shown), indicating that NF-κB activity does not act upstream of Ipf1/Pdx1 and HNF-1α to ensure high-level Glut2 expression in postnatal β-cells.

Glut2 mutant mice (26) lack first- but not second-phase GSIS. In contrast, the NIβ mice lacked both first- and second-phase insulin secretion (Figs. 4A and data not shown), indicating that the insulin secretory mechanism of the NIβ mice might also be perturbed downstream of GLUT2 expression and glucose uptake. Uncoupling protein-2 (UCP-2) impairs GSIS by uncoupling respiration from oxidative phosphorylation, which in turn leads to reduced ATP production (31,32). Overexpression of Ucp-2 in β-cells blunts GSIS (31,32), whereas both Ucp-2^+/− and Ucp-2^−/− mice show improved GSIS (32,33). However, the mechanism by which UCP-2 expression and/or activity is regulated remains largely unknown, although recent data suggest that superoxide can stimulate UCP-2 activity (34). Ucp-2 expression was increased by ∼60% in islets derived from NIβ mice when compared with control islets (Fig. 5), which is likely to impair the generation of ATP in the β-cells of NIβ mice. Collectively, these results provide evidence that impaired NF-κB activity leads, directly or indirectly, to increased Ucp-2 expression in β-cells. Thus, attenuated NF-κB activity leads to perturbed expression of key genes involved in both glucose uptake and ATP generation in β-cells.

Apart from glucose, a variety of other substances, or secretagogues, can stimulate insulin secretion from β-cells. Glybenclamide, a sulfonylurea drug, induces membrane depolarization and insulin secretion by directly blocking the SUR1/Kir6.2 K_ATP channel (35). Carbachol, a cholinergic agonist, enhances GSIS via the phospholipase C/protein kinase C signaling pathway and is suggested to increase the transport of insulin vesicles to the secretory site (1,36,37). These secretagogues depend, at least partly, on intact glucose metabolism for their ability to stimulate insulin secretion (36–42). Both secretagogues effectively stimulated insulin secretion in control mice. In contrast, NIβ mice responded poorly to glybenclamide and carbachol (Figs. 4B and C), although immunohistochemical analysis revealed normal expression of the K_ATP channel components SUR1 and the Kir6.2 subunit (data not shown). Thus, the impaired secretory responses to glybenclamide and carbachol are likely to reflect perturbed glucose uptake and utilization caused by the aberrant expression of Glut2 and Ucp-2.

Arginine stimulates insulin secretion largely independent of glucose by directly depolarizing the β-cell membrane, but it still depends on a fully functional pathway of insulin exocytosis (38,41–43). Arginine administration to fasted control mice resulted in a rapid and transient increase in serum insulin levels between 0 and 10 min after injection. The NIβ mice also responded to arginine by secreting insulin, but the secretory response was slow, and insulin levels were still elevated after 20 min (Fig. 4D). The atypical kinetic pattern of insulin secretion suggests that in addition to glucose utilization, the insulin exocytosis process is perturbed in NIβ mice.

The exact mechanism by which insulin exocytosis is regulated remains poorly defined; however, the Rab3 family of GTPases have been implicated in the control of insulin exocytosis (1). Studies in insulinoma cell lines implicate all Rab3 isoforms in nutrient-induced, but not basal, insulin secretion (44), and overexpression of both wild-type and mutant versions of the Rab3 proteins impaired Ca²⁺-triggered exocytosis in neuroendocrine cell lines (45). Interestingly, Rab3a mutant mice are glucose intolerant and show impaired insulin secretion in response to arginine. The level of expression of the Rab3C isoform (1,44) was consistently decreased by ∼80% in islets of NIβ mice (Fig. 5), and Rab3B was upregulated. No significant differences in the levels of expression of the Rab3A and D isoforms, or of the SNAP (soluble N-ethylmaleimide–sensitive factor attachment protein) receptor Vamp-2 gene, which has also been implicated in insulin exocytosis (1), were observed in islets of control and NIβ mice (Fig. 5). Taken together, the altered kinetic response to arginine and the altered expression of Rab3B and Rab3C collectively suggest that insulin exocytosis is perturbed in β-cells of the NIβ mice.

Glucose homeostasis in mammals critically depends on coordinated glucose uptake, oxidative metabolism, and insulin exocytosis in β-cells. NF-κB activity in β-cells appears to be stimulated by a rise in intracellular Ca²⁺ (8). Our results provide evidence that suppression of NF-κB activity in β-cells leads to impaired GSIS and to perturbed expression of genes involved in glucose uptake, oxidative metabolism, and Ca²⁺ triggered exocytosis of insulin. Thus, taken together, these findings suggest that NF-κB is part of a positive regulatory circuit that maintains GSIS in β-cells.

Fetal β-cells show an impaired insulin secretion in response to glucose as well as to insulin secretagogues like tolbutamide and arginine (2,3). The ability to respond to glucose and other secretagogues is significantly improved in β-cells of newborns compared with late gestational fetuses, and then improves further with age to reach the full GSIS response displayed by adult β-cells (2,3). The expression of GLUT2 correlates well with the acquirement of GSIS during β-cell maturation. GLUT2 is expressed at low levels in fetal β-cells, and the expression is then gradually increased in newborn and adult β-cells (46–48). The low-level expression of GLUT2 observed in neonatal NIβ mice, as compared with control littermates, suggests that NF-κB activity is required during fetal development and/or at birth for β-cell maturation.

Glut2 mutant mice display hyperglycemia, glucose intolerance, and a lack of first-phase insulin secretion in response to glucose (26). Patients diagnosed with Fanconi-Bickel syndrome, due to Glut2 mutations, have been reported to display diabetes, defective β-cell glucose sensing, and inappropriately low insulin secretion (49), suggesting a conserved function for Glut2 in assuring normal glucose homeostasis in both mice and humans. GLUT2 is expressed at low levels in HNF1-α–deficient mice (29,30), as well as in mice with impaired Ipf1/Pdx1 and FGFR1c activity in β-cells (27,28)—mice that all develop diabetes. Apart from GLUT2, the expression of HNF1-α, Ipf1/Pdx1, and FGFR1c signaling components, and several of their proposed target genes, was normal in NIβ mice, suggesting that NF-κB does not act upstream of these genes. Nevertheless, the impaired expression of Glut2 in Ipf1/IκBαM mice leaves open the possibility that NF-κB acts downstream of or in parallel with HNF1-α, IPF1/PDX1, and FGFR1c signaling to ensure the acquisition and maintenance of high-level Glut2 expression. NF-κB activation in response to Ca²⁺ influx appears to depend on mitogen-activated protein (MAP)/ERK (extracellular signal-related kinase) kinase (MEK)/MAP kinase (MAPK) (8) activity, and fibroblast growth factor signaling is known to activate the MEK/MAPK (50) pathway, raising the possibility that Ipf1/Pdx1 function, and hence FGFR1c signaling via MEK/MAPK, stimulates NF-κB activation to assure high-level expression of Glut2 in pancreatic β-cells.

The perturbed exocytosis of insulin displayed by NIβ mice in response to arginine may, at least in part, be explained by the downregulation of Rab3C and/or the modest upregulation of Rab3D. The exact role of Rab3 proteins in the insulin exocytotic process remains poorly defined, and neither Rab3C nor Rab3D mutant mice are available. However, Rab3A mutant mice are glucose intolerant and show a normal increase in serum insulin levels in response to a single dose of arginine, but they show a blunted response to a second consecutive arginine injection, indicating that the primary defect in Rab3A^−/− mice is at the level of recruitment of insulin vesicles to the readily releasable pool (43). In contrast, NIβ mice show a slow secretory response to a single dose of arginine, and the initial rapid burst of insulin release is absent. Together, these data suggest that the β-cells of NIβ mice can recruit insulin vesicles to the readily releasable pool, but that the final step of insulin exocytosis is perturbed. These data also suggest that despite a high degree of homology between the different Rab3 isoforms, they might be functionally different in pancreatic β-cells.

The combined perturbations in Glut2, Ucp-2, and Rab3C and D expression will perturb β-cell function at several different levels (glucose uptake, glucose metabolism, and insulin exocytosis), consequently leading to impaired GSIS. Impaired GSIS is a hallmark of type 2 diabetes (4,5), but the molecular mechanisms underlying the β-cell defects remain largely unknown. Our data provide evidence that NF-κB activation is an important step in the mechanism that ensures proper glucose-coupled stimulation of insulin secretion. Hence, the findings presented here might be relevant for our understanding of the insulin secretory defects displayed by β-cells of type 2 diabetic patients. Whether attenuation of NF-κB activity contributes to β-cell dysfunction in humans, however, remains to be determined.

This work was supported by grants from the Medical Faculty at Umeå University (to S.N.), the Swedish Research Council, the Göran Gustafsson Foundation for Research in Natural Sciences and Medicine, the Swedish Diabetes Foundation, Wallenberg Consortium North, the Juvenile Diabetes Research Foundation, and the European Union Regional Fund, Objective 1 (to H.E.).

We thank the Umeå Transgene Core Facility, C. Granberg, U. Valtersson, E. Pålsson, and I. Berglund-Dahl for technical assistance; Lena Kvist, Bo Ahrén, and U. Dahl for helpful technical instructions and suggestions; members in our laboratory for helpful discussions; and Thomas Edlund for critical reading and comments.