A Key Role in Insulin Secretion

View metadata, citation and similar papers at core.ac.uk brought to you by CORE

provided by Elsevier - Publisher Connector ARTICLE

Nicotinamide nucleotide transhydrogenase: A key role in insulin secretion

Helen Freeman,1,2 Kenju Shimomura,2 Emma Horner,1 Roger D. Cox,1,3,* and Frances M. Ashcroft2,3,*

1 Medical Research Council, Mammalian Genetics Unit, Harwell, Oxfordshire OX11 0RD, United Kingdom 2 University Laboratory of Physiology, Parks Road, Oxford, OX1 3PT, United Kingdom 3 These authors contributed equally to this work. *Correspondence: [email protected] (R.D.C.); [email protected] (F.M.A.)

Summary

The C57BL/6J mouse displays glucose intolerance and reduced insulin secretion. QTL mapping identified Nicotinamide Nu- cleotide Transhydrogenase (Nnt), a nuclear-encoded mitochondrial protein thought to be involved in free radical detoxification, as a candidate gene. To investigate its functional role, we used siRNA to knock down Nnt in insulin-secreting MIN6 cells. 2+ This produced a dramatic reduction in insulin secretion and the rise in [Ca ]i evoked by glucose, but not tolbutamide. We identified two ENU-induced point mutations in Nnt (N68K, G745D). Nnt mutant mice were glucose intolerant and secreted less insulin during a glucose tolerance test. Isolated islets showed impaired insulin secretion in response to glucose, but not to tolbutamide, and glucose failed to enhance ATP levels. Glucose utilization and production of reactive oxygen species were increased in Nnt b cells. We hypothesize that Nnt mutations/deletion uncouple b cell mitochondrial metabolism leading to less ATP production, enhanced KATP channel activity, and consequently impaired insulin secretion.

Introduction 2001). Changes in b cell metabolism are coupled to changes in insulin release via metabolic regulation of the activity of the Type 2 diabetes mellitus is a serious metabolic disease that is ATP-sensitive potassium (KATP) channel (Ashcroft and Rorsman reaching epidemic proportions in Western societies and is pre- 1989, 2004)(Figure 1). At low blood glucose levels, KATP chan- dicted to affect 300 million people worldwide by 2025 (King nels are open, hyperpolarizing the b cell membrane and prevent- et al., 1998). Impaired glucose tolerance, which precedes diabe- ing Ca2+ influx and insulin secretion. When glucose levels rise, tes and is a risk factor for the disease, currently affects a further increased b cell metabolism produces elevation of cellular 200 million worldwide. Type 2 diabetes is characterized by ele- ATP levels. This causes closure of KATP channels, which leads vation of the blood glucose concentration, usually presents in to membrane depolarization, activation of voltage-gated Ca2+ 2+ 2+ middle age, and is exacerbated by age and obesity. It is associ- channels, Ca entry, and a rise in [Ca ]i that triggers insulin ated with both impaired insulin secretion and insulin action, but it secretion. is now recognized that b cell dysfunction is a key element in the Earlier studies have shown that the reduced insulin secretion development of the disease (Bell and Polonsky, 2001; Kahn, of C57BL/6J mice derives from the failure of glucose to close 2003; Ashcroft and Rorsman, 2004). pancreatic b cell KATP channels (Toye et al., 2005). Conse- Mouse models of diabetes and glucose intolerance may pro- quently, the membrane potential is not depolarized and Ca2+ en- vide useful insights into insulin secretory mechanisms and the try is prevented. However, the sensitivity of KATP channels in etiology of human type 2 diabetes. One such example is the C57BL/6J mice to inhibition by ATP (IC50 =10mM; Toye et al., C57BL/6J mouse strain, which shows impaired glucose toler- 2005) is similar to that observed for NMRI mice (Ashcroft and ance under free-fed and postprandial conditions: using an intra- Kakei, 1989) and for cloned b cell KATP channels expressed in peritoneal glucose tolerance test (IPGTT), the blood glucose heterologous systems (Tucker et al., 1997). This suggests that level of nonobese C57BL/6J mice rises higher and takes longer the defect in C57BL/6J mice lies prior to KATP channel closure, to regain the resting level than that of other mouse strains (Kaku probably at the level of b cell metabolism (Toye et al., 2005). et al., 1988; Rossmeisl et al., 2003; Kulkarni et al., 2003; Toye By intercrossing nonobese C57BL/6J and C3H/HeH mice and et al., 2005). This is associated with reduced plasma insulin lev- employing QTL mapping, we identified three main genetic loci els, resulting from impaired first and second phase insulin re- that influence the difference in glucose homeostasis (Toye lease (Kaku et al., 1988; Rossmeisl et al., 2003; Kooptiwut et al., 2005). These were located on chromosomes 9, 11, and et al., 2002; KayoRETRACTED et al., 2000). These effects are magnified in 13. Within the chromosome 13 locus lay the gene encoding Nic- mice that possess a genetic defect in insulin resistance (Kulkarni otinamide Nucleotide Transhydrogenase (Nnt), and sequencing et al., 2003). In addition, C57BL/6J mice develop obesity, hyper- of the coding region of Nnt in C57BL/6J mice identified two mu- glycemia, and insulin resistance when fed a high-fat diet (Ross- tations (Toye et al., 2005). The first is a missense mutation meisl et al., 2003). (M35T) in the mitochondrial leader sequence of the Nnt precur- Insulin secretion from the b cells of the pancreatic islets is crit- sor protein. The second is an in-frame 5-exon deletion that re- ically dependent on cellular ATP levels generated by glucose- moves four predicted transmembrane helices and their con- stimulated mitochondrial metabolism (Maechler and Wollheim, necting linkers.

CELL METABOLISM 3, 35–45, JANUARY 2006 ª2006 ELSEVIER INC. DOI 10.1016/j.cmet.2005.10.008 35 ARTICLE

compared to wild-type b cells. Our data are consistent with the idea that impaired Nnt function reduces b cell metabolism, thereby preventing closure of KATP channels in response to glucose, calcium inﬂux, and insulin secretion.

Results

siRNA knockdown of Nnt in MIN6 cells To determine the functional role of Nnt in b cells, we used siRNA to silence Nnt in the insulin-secreting b cell line MIN6. MIN6 cells were mock-transfected or transfected with one of four different siRNA duplexes: nonsense, glucokinase (Gck), or two different Nnt-specific siRNAs. Gck was used as positive control, as it is a key glycolytic enzyme, and in its absence, insulin secretion is abolished (Terauchi et al., 1995). A scrambled (nonsense) Figure 1. Mechanism of insulin secretion from pancreatic b cells siRNA was used as a negative control. We used two separate Under resting conditions, KATP channels are open, producing hyperpolarization of the b cell plasma membrane and so preventing insulin secretion. Glucose uptake Nnt siRNAs targeted to different regions of the gene. All siRNAs (via the GLUT2 transporter) and its subsequent metabolism elevates ATP and re- were tagged with a fluorescent probe to enable transfected cells duces MgADP levels, causing KATP channels to close. This produces a membrane to be easily detected. depolarization that activates voltage-gated Ca2+ channels, causing Ca2+ influx and 2+ Quantitative real-time PCR and Western blotting were used to a rise in cytosolic Ca that triggers release of insulin granules. ATP also stimulates verify mRNA and protein knockdown. Expression of Gck and the secretory machinery directly. Nnt mRNAs was abolished by siRNAs targeted against these genes, but neither was affected in cells transfected with non- Nnt is a nuclear-encoded mitochondrial protein that is located sense siRNA (Figure 2A). Similar results were found for protein in the inner mitochondrial membrane of eukaryotes and in the expression. Figure 2B shows Western blots of MIN6 cells plasma membrane of some bacteria (for review, see Hoek and probed with antibodies to Gck and Nnt following transfection Rydstrom, 1988). It functions as a redox-driven proton pump, with one of the three different siRNA. Islets isolated from C3H/ catalyzing the reversible reduction of NADP+ by NADH and con- HeH and C57BL/6J mice are also included as positive and neg- version of NADH to NAD+. This is linked to proton translocation ative controls, respectively. The Nnt protein is not found in across the membrane according to the reaction: C57BL/6J mice, although as previously reported, the RNA is present (Toye et al., 2005). It is also clear that Nnt is present in + + MIN6 cells transfected with nonsense (upper panels) or Gck NADH + NADP + Hcytosolic4NAD + NADPH + Hmatrix siRNA (lower panels) but not in cells transfected with Nnt siRNA Under physiological conditions, a marked proton gradient is present (middle panels). Transfection with Gck siRNA abolished expres- across the inner mitochondrial membrane so that the reaction is sion of Gck without altering Nnt expression (Figure 2B, lower driven strongly to the right. This makes Nnt a very efficient generator panels). A second Nnt siRNA also abolished protein and RNA of NADPH. Moreover, it couples the production of NADPH to the expression (data not shown). Thus, we achieved target-specific rate of mitochondrial metabolism, and hence to the production of knockdown of Nnt with both siRNAs. reactive oxygen species (ROS) generated by the electron transport chain. Thus, Nnt is postulated to play an important role in the detox- Insulin secretion in MIN6 cells ification of ROS (Hoek and Rydstrom, 1988; Arkblad et al., 2005). We next compared insulin secretion from MIN6 cells transfected Nnt is a homodimer and the individual subunits (Mr w110 kDa) with nonsense siRNA or Nnt siRNA. Figure 2C shows that basal are composed of three domains (Hoek and Rydstrom, 1988). insulin secretion in glucose-free solution was almost identical for The N-terminal hydrophilic domain (I) binds NAD(H), the central both nonsense and Nnt siRNA-transfected cells but that glucose hydrophobic domain (II) intercalates into the membrane (it con- failed to stimulate insulin secretion from cells in which Nnt had sists of 14 transmembrane domains and forms the proton chan- been knocked down. However, the sulphonylurea tolbutamide nel), and the C-terminal hydrophilic domain (III) binds NADP(H) (200 mM) elicited equivalent responses in both nonsense and (Yamaguchi and Hatefi, 1991). Both domains I and III are located Nnt siRNA-transfected cells. Because tolbutamide closes KATP within the mitochondrial matrix. The 5-exon deletion found in the channels directly, this result suggests that the defect in insulin Nnt gene of C57BL/6J mice deletes part of domain I and the first secretion lies prior to KATP channel closure. Under our con- four transmembrane domains and leads to a marked downregu- ditions, about 95% of cells were transfected, as evidenced lation of Nnt mRNA in liver and islets (Toye et al., 2005). by their fluorescence. This explains why knockdown of Nnt In this paper, we explore the functional role of Nnt in detail. We was able to almost completely abolish insulin secretion in our show that knockdownRETRACTED of Nnt in insulin-secreting MIN6 cells pre- preparation. vents elevation of intracellular calcium and insulin secretion in response to glucose, but not to the KATP channel blocker tolbu- Calcium imaging in MIN6 cells tamide. Likewise, mice carrying mutations in Nnt have impaired To determine the cause of impaired insulin secretion, we mea- glucose tolerance and loss of both glucose-dependent insulin sured changes in the intracellular calcium concentration 2+ secretion and ATP production in isolated pancreatic islets. In ([Ca ]i) of MIN6 cells evoked by glucose. In mock-transfected addition, we found that mutant Nnt b cells show enhanced glu- cells, glucose concentrations greater than 2 mM elicited a rise 2+ cose utilization and ROS production in response to glucose, in [Ca ]i (Figures 3A and 3C). This response was unaltered by

36 CELL METABOLISM : JANUARY 2006 Nnt and insulin secretion

Figure 2. Effect of Nnt knockdown on insulin release from MIN6 cells A) Comparative levels of Gck (top) and Nnt (bottom) gene expression using quantitative PCR in MIN6 cells transfected with nonsense siRNA (ns control, gray bar), Nnt siRNA (white bar), and Gck siRNA (black bar). The data are the mean 6 SD of ﬁve duplicates and are representative of three separate experiments. B) Western blots of C3H/HeH (C3H) or C57BL/6J (B6) mouse islets (left two lanes) and of MIN6 cells (right lanes) transfected with nonsense siRNA (top), Nnt siRNA (middle), or Gck siRNA (bottom). NS, nonsense-transfected MIN6 cells for comparison. Num- bered lanes, number of MIN6 repeats of transfected cells. Each blot was probed with three different antibodies (Ab), as indicated on the left, against Nnt, Gck, and actin (loading control, data not shown). The data are representative of seven separate experiments. C) Insulin secretion in response to glucose (0, 2, 10 mM) or tolbutamide (200 mM) from MIN6 cells transfected with nonsense siRNA (white bars, n = 6) or with Nnt siRNA (black bars, n = 6). The data are the mean 6 SEM of six cells and are representative of three separate experiments.

transfection with nonsense siRNA but was abolished by Gck 745. Both residues N68 and G745 are highly conserved across siRNA and markedly reduced by both Nnt siRNAs (Figures 3B species (Figure 4A). and 3C). These results resemble those found for C57BL/6J Both mutants were recovered as live mice by IVF (male F1 mice (Toye et al., 2005) and demonstrate that the failure of glu- BALB/c 3 C3H/HeH sperm and C3H/HeH eggs), and the prog- cose to induce insulin secretion produced by downregulation of eny were intercrossed to produce mice homozygous for the mu- Nnt is due to the inability of glucose to promote Ca2+ influx. Tol- tations. Intraperitoneal glucose tolerance tests (IPGTT) revealed butamide (200mM), however, stimulated a similar increase in that mice heterozygous and homozygous for the N68K mutation 2+ [Ca ]i in both cells transfected with nonsense and Nnt siRNA were markedly less glucose tolerant than their wild-type litter- (Figure 3D). The simplest explanation of these results is that glu- mates (Figure 5A). During a 2 hr IPGTT, blood glucose levels cose fails to close KATP channels when Nnt is down regulated. rose significantly higher and took longer to return to baseline in homozygous 12-week-old male mice. Furthermore, heterozygotes were not as glucose intolerant as homozygotes (Figure Nnt mutant mice show glucose intolerance 5A). These differences in glucose tolerance persisted at 16 and reduced insulin secretion weeks (Figure 5C). Both heterozygous and homozygous N68K We identified two mutant alleles of Nnt by screening the Harwell mice also secreted substantially less insulin over a 30 min period ENU-DNA archiveRETRACTED (Quwailid et al., 2004). The first (N68K) was an following a glucose challenge (Figure 5E). Very similar results exchange of C for A (nt 204) in exon 2 that results in a nonconser- were obtained for female mice (see Supplemental Data available vative missense asparagine to lysine mutation at amino acid 68 with this article online). of the protein. This mutation is located in the predicted NAD Figure 5B shows that 12-week-old male mice heterozygous binding domain of Nnt. The second mutation (G745D) was found and homozygous for the G745D mutation were also significantly in the membrane spanning domain of Nnt and consisted of an less glucose tolerant than controls during a 2 hr IPGTT. Interest- exchange of G for A (nt 2224) in exon 14, resulting in a noncon- ingly, the glucose tolerance profiles over 2 hr are very similar servative substitution of glycine by aspartic acid at amino acid for heterozygous and homozygous mice, suggesting that the

CELL METABOLISM : JANUARY 2006 37 ARTICLE

Figure 3. Effect of Nnt knockdown on Ca2+ transi- ents in MIN6 cells A) Representative changes in intracellular Ca2+, measured by Fura-2 fluorescence, in response to glucose and tolbutamide in single MIN6 cells. Bars indicate when glucose (2, 5, 10 mM glucose) or tolbutamide (200 mM, Tol) were applied. Each color repre- sents a different cell. 2+ B) Representative changes in [Ca ]i, measured by Fura-2 fluorescence, in response to glucose and tolbutamide in single MIN6 cells transfected with Cy3- conjugated siRNA duplexes for NntI or NntII. Bars indicate when glucose (2, 5, 10 mM glucose) or tolbutamide (200 mM, Tol) were applied. Each color repre- sents a different cell. 2+ C) Mean change in [Ca ]i, in response to 2, 5, and 10 2+ mM glucose, normalized to the resting [Ca ]i,in 0 mM glucose for mock-transfected (orange) MIN6 cells or MIN6 cells transfected with nonsense (blue), NntI (green), NntII (cyan), or Gck (pink) siRNA. Each data point is the mean 6 SEM of 15–21 cells. The data are representative of five separate experiments. 2+ D) Mean change in [Ca ]i, in response to 200 mMtol- butamide for MIN6 cells mock-transfected (orange) or transfected with nonsense (blue), NntI (green), NntII (cyan) or Gck (pink) siRNA. The data indicate the mean 6 SEM of 15-21 cells. The data are representative of 5 separate experiments.

G745D mutation affects heterozygous mice as severely as ho- lets. Background secretion at 0 or 2mM glucose from homozy- mozygous mice. Insulin secretion and glucose tolerance over gous or heterozygous G745D islets was not significantly 30 min were also impaired in 16-week-old mice (Figures 5D different from that of wild-type islets, but glucose (10 mM) stim- and 5F). Similar results were found for G745D mutant female ulated insulin secretion was substantially reduced. However, the mice (see Supplemental Data). Thus, two separate mutations ability of tolbutamide to stimulate secretion was similar in all in the Nnt gene confer impaired glucose tolerance and insulin three types of mice (Figure 6A). Similar results were obtained responses. No difference in thermoregulation was observed for for islets isolated from heterozygous or homozygous N68K either Nnt mutation (see Supplemental Data). mice (data not shown). These results resemble those found for MIN6 cells and suggest that both Nnt mutations produce Nnt expression in Nnt mice a loss of protein function. Western blotting was used to determine if Nnt mutant mice ex- press the Nnt protein. Figure 4B shows that Nnt was expressed ATP measurements in Nnt mice islets in islets of homozygous N68K and G745D Nnt mice; therefore, The data reported here, and those found for C57BL/6J mice these mutations mustRETRACTED affect protein function and/or mitochon- (Toye et al., 2005), are consistent with the idea that Nnt deletion drial membrane targeting. or mutation impairs b cell metabolism and thereby insulin secretion. We therefore measured ATP concentration in islets isolated Insulin secretion in Nnt mice islets from wild-type, heterozygous and homozygous Nnt mutant We next measured insulin secretion from islets isolated from Nnt mice. As Figure 6B shows, ATP content increased significantly mutant mice. Insulin secretion was normalized to insulin content (1.8-fold; p < 0.01) in wild-type islets when glucose was in- to control for any differences in islet size. There was no signifi- creased from 2–20 mM. However, no increase was detected cant difference in insulin content between wild-type or mutant is- in either heterozygous or homozygous G745D Nnt islets (Figure

38 CELL METABOLISM : JANUARY 2006 Nnt and insulin secretion

Figure 4. Identiﬁcation of Nnt mutations A) Alignment of Nnt amino acid sequence for various species across the regions in which mutations in Nnt were identiﬁed by screening the Harwell ENU-DNA archive. The mutated amino acids in exon 2 (N68K) and exon 14 (G745D) are indicated. B) Western blots of islets isolated from homozygous N68K or G745D Nnt mutant mice. Each blot was probed with antibodies (Ab) against Nnt and actin, as indicated. The data are representative of four separate experiments.

6B). No significant difference was observed in resting ATP con- phonylurea tolbutamide is used to close KATP channels. These tent (at 2 mM glucose). Similar results were obtained for N68K results are consistent with the idea that metabolic regulation islets (data not shown). of KATP channel activity is impaired in Nnt mutant mice, or following Nnt knockdown in the insulin-secreting MIN6 cell line. Reactive oxygen species measurements They are also harmonious with the location of Nnt in the inner mi- Several studies have suggested that Nnt is involved in detoxifi- tochondrial membrane. cation of reactive oxygen species (ROS) generated by the electron transport chain (reviewed by Hoek and Rydstrom, 1988). Role of Nnt We therefore measured ROS production in single b cells isolated Two functional roles have been proposed for Nnt: the detoxifica- from wild-type and homozygous Nnt G745D mice, using the tion of reactive oxygen species (Hoek and Rydstrom, 1988; Ark- fluorescent dye 20,7’-dichlorofluorescein (Figure 6C). ROS pro- blad et al., 2005) and regulation of isocitrate dehydrogenase ac- duction was not significantly different in the absence of glucose tivity (Sazanov and Jackson, 1994). Our results suggest that the between wild-type and mutant b cells. Glucose (20 mM) pro- former may be of primary importance for the mechanism by duced a small increase (3-fold) in ROS production in wild-type which loss of Nnt function impairs glucose-dependent ATP pro- b cells, but a far more dramatic (w10-fold) increase in Nnt duction and insulin secretion. G745D mouse b cells (Figure 6C). Reactive oxygen species (such as superoxide) are produced by electron leakage during mitochondrial metabolism primarily Glucose utilization by Nnt mice islets at Complexes I and III (Raha and Robinson, 2000). A significant It has been shown that ROS increase the activity of uncoupling fraction (at least 50%) of the superoxide generated is released proteins, such as UCP2 (Echtay et al., 2002; Krauss et al., 2003; into the mitochondrial matrix (Muller et al., 2004) and as it is Brand et al., 2004). This leads to enhanced proton leakage membrane impermeant must be detoxified within the mitochon- across the inner mitochondrial membrane, which reduces the drion itself. This is achieved by conversion to H2O2 by manga- electromotive force and thereby ATP generation. Thus a possi- nese-dependent superoxide dismutase 2 (SOD2) and subse- ble explanation for the lower ATP production of Nnt G745D quent oxidation to water by the glutathione cycle (Figure 7). mouse islets observed in 20 mM glucose is that it results from This process requires a continuous supply of NADPH. In mito- mitochondrial uncoupling elicited by the greater levels of ROS. chondria, this is provided by nicotinamide nucleotide transhy- To test this idea, we measured glucose utilization. As Figure drogenase, which converts NADP+ and NADH into NADPH 6D shows, glucose utilization in response to 20 mM glucose and NAD+. Thus, in the absence of Nnt, one might expect an in- was greater in islets isolated from homozygous Nnt G745D crease in the concentration of H2O2 and other reactive oxygen mice than wild-type mice, whereas ATP production was lower. species. This result is characteristic of an uncoupler (Green et al., 2004) We observed that glucose stimulated a small increase in ROS and suggests that at high glucose levels, mitochondria in Nnt in wild-type b cells. This is consistent with multiple previous G745D islets are functionally uncoupled. studies showing that the rate of ROS production is enhanced as metabolism increases (Krauss et al., 2003; Bindokas et al., Discussion 2003; Fridlyand and Philipson, 2004), although this has been contested (Martens et al., 2005). Importantly, we found that The results described here demonstrate that Nnt plays an im- ROS production was greatly enhanced in Nnt mutant b cells, portant role in the regulation of insulin secretion from the pan- consistent with a role for Nnt in ROS detoxification. creatic b cell. MiceRETRACTED in which this protein is deleted (e.g., Several studies have shown that ROS, such as H2O2, inhibit C57BL/6J) or mutated (Nnt-N68K and Nnt-G745D), are charac- insulin secretion (Maechler et al., 1999; Krippeit-Drews et al., terized by impaired glucose tolerance. This is a consequence of 1999). This results from depolarization of the b cell mitochondrial impaired ATP production in response to glucose elevation, membrane potential, which inhibits ATP production and pro- which translates into a marked decrease in insulin secretion. motes activation of KATP channel currents, plasma membrane Furthermore, siRNA silencing of Nnt in insulin-secreting MIN6 hyperpolarization, and inhibition of secretion (Maechler et al., 2+ cells abolishes the glucose-dependent increase in both [Ca ]i 1999; Krippeit-Drews et al., 1999). The mechanism by which and insulin secretion. However, both are elicited when the sul- ROS influence the mitochondrial membrane potential and ATP

CELL METABOLISM : JANUARY 2006 39 ARTICLE

Figure 5. Plasma glucose and insulin dynamics in wild-type and Nnt mutant mice A and B) Plasma glucose dynamics measured in response to an intraperitoneal glucose tolerance test (IPGTT) following overnight fasting for 12-week-old male mice. The IPGTT was carried out over 2 hr. Data points indicate the mean 6 SD. (A) wild-type (circles, n = 9), heterozygous (squares, n = 10), and homozygous (triangles, n = 15) mice carrying the N68K mutation in Nnt.(B) wild-type (circles, n = 10), heterozygous (squares, n = 16), and homozygous (triangles, n = 13) mice carrying the G745D mutation in Nnt. Statistical signiﬁcance of homozygote versus wild-type and heterozygotes versus wild- type is indicated, *p < 0.05 and **p < 0.01 (Student’s t test). C–F) Plasma glucose (C and D) and insulin (E and F) dynamics measured in response to an IPGTT at 16 weeks for the same mice as in (A and B). The IPGTT was carried out over 30 min. For key, see (A and B). Data points indicate the mean 6 SD.

production is believed to involve activation of uncoupling protein mice fail to upregulate antioxidant enzymes in islets in response 2 (UCP2), since UCP2 activity is stimulated by ROS (Echtay to oxidative stress and show more severely impaired glucose et al., 2002; Krauss et al., 2003; Brand et al., 2004). This leads tolerance than female mice (Friesen et al., 2004). to enhanced proton leakage across the inner mitochondrial It has also been suggested that Nnt may exert a fine control of membrane, reducing the electromotive force and thereby ATP flux through the tricarboxylic acid cycle, by regulating the activ- production. Consequently, increased UCP2 activity impairs ity of isocitrate dehydrogenase (ICD) (Sazanov and Jackson, insulin secretion (Krauss et al., 2003; Chan et al., 2001) and 1994). This is because NAD+ (produced by Nnt) is required for reduced UCP2 activity increases insulin release (Zhang et al., the activity of ICD. However, this mechanism seems less likely 2001). to be primarily responsible for the effects of Nnt knockdown, The combination of increased glucose utilization and reduced since reduced NAD+ production would be predicted to impair ATP production found for Nnt G745D islets is consistent with the glucose metabolism, whereas we observed glucose utilization idea that their b cell mitochondria become increasingly un- was enhanced. coupled when glucose is elevated. This is predicted if the increased ROS production we observed at high glucose stimu- Nnt ENU mutant mice lates UCP2. We therefore propose that the inability of glucose Both of our mutant mouse models of Nnt were significantly glu- to elevate ATP in Nnt mutant islets is mediated by ROS-induced cose intolerant and had impaired insulin secretion. In this re- activation of UCP2 (Figure 7). Because glucose does not in- spect, the phenotype was similar to that found for C57BL/6J crease ATP levels,RETRACTED KATP channels remain open, keeping the mice, which have a 5-exon deletion in Nnt (Toye et al., 2005). membrane hyperpolarized and inhibiting insulin secretion. ENU mutagenesis introduces multiple mutations in the genome Since more than 95% of ATP in the b cell is produced by the and with a per locus mutation rate of 0.00108 and 30,000 genes, mitochondria, and mitochondrially generated ATP is of primary there will be an estimated 32 dominant functional mutations per importance for insulin release (Silva et al., 2000); Maechler and F1 founder mouse (Quwailid et al., 2004; Coghill et al., 2002). Wollheim, 2001), small changes in ROS detoxification by Nnt These mutations are segregated by subsequent backcrossing might have a significant effect on insulin secretion. This idea is and, ultimately, intercrossing. The two ENU Nnt mutations we consistent with the fact that male (but not female) C57BL/6J identified were independently derived and are thus extremely

40 CELL METABOLISM : JANUARY 2006 Nnt and insulin secretion

Figure 6. Effects of Nnt mutations on properties of islet cells A) Insulin secretion from islets isolated from G745D Nnt mice in response to glucose (0, 2, 10 mM) or tolbutamide (tol, 200 mM). Islets were isolated from two mice of each type for these experiments. The data indicate the mean 6 SD of six replicates (each of five islets) and are representative of two separate experiments on G745D mice and two on N68K mice. Black bars represent homozygous mutants, hatched bars represent heterozygotes, and white bars represent wild-type littermates. B) ATP content of islets isolated from G745D Nnt mice in response to glucose (2 mM gray bars, 20 mM black bars). Data are normalized to protein content. The data indicate the mean 6 SEM of five replicates (each of six islets) and are representative of two separate experiments on G745D mice. C) ROS measured using the fluorescent dye DCF in single b cells isolated from wild-type mice (white bars) and homozygous Nnt G745D mutant mice (black bars). Data indicate the mean 6 SEM of 30–50 cells at each glucose concentration (pooled data from three separate experiments). Inset: representative images of DCF fluoresecence in b cells isolated from wild-type and G745D mutant mice incubated in 0 or 20 mM glucose. D) Glucose utilization measured at different glucose concentrations for wild-type mice (white bars) and homozygous G745D Nnt mutant mice (black bars).The data indicate the mean 6 SEM of seven duplicates (each of five islets) and are representative of two separate experiments. unlikely to share any ENU-induced mutations in genes other mate the contribution of an individual gene to an individual quan- than Nnt. Furthermore, neither allele is found in the parental titative trait because the effect is estimated in a genetically strains. Thus, the observed phenotypes are due to the Nnt mu- heterogeneous F2 population using linked genetic markers. tations carried by the two lines. Furthermore, in an F2 population, new gene interactions may be Both homozygous Nnt mutations appear to have a greater ef- unmasked (Stoehr et al., 2000). Thus, the phenotype of the F2 fect on glucose tolerance than that observed for C57BL/6J generation may differ from that observed on the original (or in- mice, yet QTL mapping suggested that Nnt contributed <10% deed other) genetic backgrounds. This may explain why Nnt of the glucose intolerant phenotype of C57BL/6J mice (Toye and C57BL/6J mice have a similar phenotype despite QTL map- et al., 2005). The ENU mutations are on a BALB/c background ping suggesting Nnt is only a minor contributor to the C57BL/6J and have been backcrossed to C3H/HeH and then intercrossed phenotype. to produce homozygous animals. In contrast, the QTL mapping There was no significant difference in phenotype between the was performed between C57BL/6J and C3H/HeH in an F2 inter- heterozygous and homozygous Nnt mice for the G745D mutation. cross. Thus the geneticRETRACTED composition of the mice is different. The Likewise, there was no difference in ATP generation or insulin smaller variance seen in the QTL mapping studies could be ex- secretion between heterozygous and homozygous animals. Be- plained if C57BL/6J mice possessed compensatory mecha- cause Nnt functions as a homodimer (Hoek and Rydstrom, nisms that reduced the severity of the Nnt deletion. The effect 1988), this might be due to a dominant-negative effect in which of genetic background on phenotype is well established: one the presence of a mutant subunit in the homodimer impairs the example is the ob/ob mutation that shows large differences in function of the other subunit, thus inactivating the whole dimer. Al- diabetes susceptibility on BTBR and C57BL/6J genetic back- ternatively it may be due to a gene dosage effect (i.e., >50% of grounds (Stoehr et al., 2000). QTL mapping may also underesti- functional protein is required to produce the wild-type phenotype).

CELL METABOLISM : JANUARY 2006 41 ARTICLE

Figure 7. Suggested mechanism for the effects of Nnt mutations The activity of the electron transport chain leads to a proton gradient across the inner mitochondrial membrane that is used to drive the production of ATP by ATP synthase. Consequently, proton leakage through uncoupling protein 2 (UCP2) decreases ATP production. UCP2 is activated by reactive oxygen species (ROS), suchasO2 , generated by the electron transport chain. In b cell mitochondria of control mice, ROS are ﬁrst reduced to H2O2 by mitochondrial superoxide dismutatase (SOD2), and then to water, using reducing equivalents supplied by the glutathione couple (GSSG-GSH) and the reduction of NADP+ to NADPH by nictotinamide nucleotide transhydrogenase (NNT). Loss of NNT function is therefore predicted to increase ROS, activate UCP2, and so reduce ATP production and insulin secretion.

Why is the defect only seen in b cells? This may be partially a consequence of reduced mRNA expres- Although Nnt is widely expressed (Arkblad et al., 2002), it ap- sion of GLUT1, GLUT2, and glucokinase. However, an increase pears that only the b cell is particularly sensitive to mutation in markers of oxidative stress, which correlates with the degree deletion of the Nnt gene as mice carrying Nnt mutations of impairment in glucose-stimulated insulin release, has also do not suffer from any obvious problems other than glucose in- been observed. Interestingly, following 24 hr culture in the pres- tolerance. Strikingly the C57BL/6J strain has been widely used ence of 30 mM glutathione, the secretory response of type 2 di- for decades without the deficiency in Nnt being phenotypically abetic islets to glucose improved significantly (Del Guerra et al., obvious (although, of course, we do not know when it arose in 2005). Thus an underlying metabolic defect, resulting in in- the stock). The lack of an extra-pancreatic phenotype is not be- creased oxidative stress, may contribute to impaired insulin se- cause of redundancy, since there is only one copy of the gene cretion in human type 2 diabetic patients (see also Green et al., in the genome and no other protein has been proposed to serve 2004). Given the effects of impaired Nnt function in mice, it a similar role. A possible explanation may be that the b cell is would clearly now be worth screening human patients to deter- particularly sensitive to small changes in cytosolic ATP, per- mine if polymorphisms in the Nnt gene are associated with type haps because its metabolism has evolved to be very sensitive 2 diabetes. Furthermore, the key role of Nnt in ROS detoxifica- to blood glucose levels. Furthermore, the b cell resting potential tion, and the fact that insulin secretion is abolished when this is largely determined by the activity of the KATP channel (Ash- gene is mutated or deleted, suggests that drugs which enhance croft and Rorsman, 1989, 2004). Small changes in KATP cur- Nnt function or expression might provide a valuable new strat- rents may have marked effects on insulin secretion, as evi- egy for treating type 2 diabetes. denced by the effects of mutations in Kir6.2, the pore-forming subunit of this channel (Sakura et al., 1995), in both human pa- Experimental procedures tients with neonatal diabetes (Gloyn et al., 2004; Proks et al., 2004) and in transgenic mice (Koster et al., 2000). Although Animal husbandry Kir6.2 is expressed in multiple tissues (Sakura et al., 1995), ex- Mice were kept in accordance with UK Home Office welfare guidelines and cept for the most severe mutations, no functional effects are project license restrictions under controlled light (12 hr light and 12 hr dark manifest in tissuesRETRACTED other than the b cell. Thus the effects of cycle), temperature (21ºC 6 2ºC) and humidity (55% 6 10%) conditions. small changes in ATP production (rather than ATP sensitivity) They had free access to water (25 ppm chlorine) and were fed ad libitum on a commercial diet (SDS maintenance chow) containing 2.6% saturated produced by impaired Nnt function may be largely confined fat. to the b cell.

Genetic crosses Physiological relevance for type 2 diabetes The Harwell ENU-DNA archive (Quwailid et al., 2004) was screened and two Compared to control islets, type 2 diabetic islets have reduced ENU-induced point mutations on Nnt were identiﬁed by denaturing high per- glucose oxidation and insulin secretion (Del Guerra et al., 2005). formance liquid chromatography (dHPLC) using a Transgenomic WAVE

42 CELL METABOLISM : JANUARY 2006 Nnt and insulin secretion

system. These mice were rederived by IVF from frozen sperm using C3H/ 1 hr. Membranes were probed with a custom-made polyclonal Nnt antibody HeH eggs, and the progeny intercrossed to produce mice homozygous for (raised against peptides, Eurogentec), Gck antibody, and actin antibody each mutation. Genomic DNA was extracted from the mouse tail tissue using (control) from Santa Cruz. ECL plus (Amersham) was then used according a Qiagen DNeasy tissue kit. Mice were genotyped by pyrosequencing. Pri- to manufacturer’s instructions to allow visualization of protein on ECL film, mers were designed from published records (Arkblad et al., 2001) using Py- developed using the Compact X4 (Xograph). rosequencing Assay Design software from Biotage. The pyrosequencing results were analyzed using the Pyrosequencer PSQ HS 96A system. Fura-2 calcium imaging MIN6 cells were cultured on 35 mm Fluorodishes (World Precision Instru- Intraperitoneal glucose tolerance tests ments) and incubated with DMEM plus 3 mM Fura-2-AM (Molecular Probes) Mice were fasted overnight, weighed, and a blood sample collected by tail for 40 min at 37ºC. They were imaged at room temperature (20ºC–24ºC) us- venipuncture under local anesthetic (Lignocaine) using Lithium-Heparin mi- ing an IonOptix fluorescence system (Boston, Massachusetts), with 340 nm crovette tubes (Sarstedt) to establish a baseline glucose level ‘‘T0’’. They and 380 nm dual excitation. The 510 nm emission ratio was collected at 1 Hz. were then injected intraperitoneally with 2 g glucose (20% glucose in 0.9% Background subtraction was performed by measuring fluorescence from NaCl)/kg body weight and blood samples taken at the designated times. a cell-free region in the field of view. Cells were perfused continuously with For a fuller protocol, see EMPReSS (the European Mouse Phenotyping Re- extracellular solution containing (mM): 137 NaCl, 5.6 KCl, 2.6 CaCl2, 1.2 source for Standardized Screens designed by Eumorphia) simplified IPGTT MgCl2, 10 HEPES (pH 7.4 with NaOH) plus glucose or tolbutamide as indi- (http://empress.har.mrc.ac.uk). Plasma glucose was measured using an cated. Only data from cells that responded for 30 min prior to addition of Analox Glucose analyser GM9. Plasma insulin was measured using a Mer- 500 mM tolbutamide were analyzed. codia Ultra-Sensitive Mouse ELISA kit according to the manufacturer’s instructions. Insulin secretion MIN6 cells (1 3 105) were replated, transfected with various siRNA’s, and cul- MIN6 cell culture tured for 24 hr (as above) before being used in secretion assays. Islets were MIN6 cells (starting passage 40) were maintained in DMEM containing isolated from wild-type and mutant mice and cultured overnight in PIM 25 mM glucose and supplemented with 15% heat-inactivated fetal bovine medium (l-cell technology, Nevada) before insulin secretion was assessed. serum in humidified 7% CO , 95% air at 37ºC. Cells were exposed to 2 Insulin secretion was measured during 1 hr static incubations in Krebs- glucose-free extracellular solution for 30 min prior to measurement of insulin Ringer Buffer (KRB solution; in mM: 118.5 NaCl, 2.54 CaCl , 1.19 KH PO , secretion or intracellular calcium. 2 2 4 4.74 KCl, 25 NaHCO3, 1.19 MgSO4, 10 HEPES, [pH 7.4]) containing 0, 2, or 10 mM glucose or 0.2 mM tolbutamide. Each glucose concentration was rep- Islet and b cell isolation licated five to eight times. Samples of the supernatant were assayed for Mice were killed by cervical dislocation, the pancreas removed, and islets insulin. To determine total insulin content, insulin was extracted using 95:5 isolated by liberase digestion and handpicking (Toye et al., 2005). Isolated is- ethanol/acetic acid. Insulin was measured using a Mouse Insulin ELISA kit lets were dispersed into single cells by incubation in calcium-free Hank’s (Mercodia, Sweden). solution (1 mM EGTA) and trituration in RPMI tissue culture medium containing 5.5 mM glucose (Gibco, UK) supplemented with 10% fetal calf serum, 100 U/ml penicillin, and 100mg/ml streptomycin. Cells were maintained in this ATP measurements Batches of six islets were incubated at 37ºCfor1hrin50ml KRB solution con- medium at 37ºC in a humidified atmosphere at 5% CO2 in air, and used 1–2 days after the isolation. taining 2 mg/ml BSA and various concentrations of glucose. After this time, 25 ml of ice-cold 10% perchloric acid was added and islets were disrupted by vortexing for 10 s. Each glucose concentration was tested in five separate siRNA assays, and ATP was measured in duplicate in 10 ml aliquots from each as- Synthetic siRNAs were designed by Eurogentec and synthesized by both say, using a luciferin-luciferase bioluminescent assay (Sigma St Louis, Mis- Eurogentec and VH Bio Ltd. A total of four siRNAs were used; Nonsense souri) and a TD20/20 luminometer (Turner Designs, Sunnyvale, California). (NS), Glucokinase (Gck), and two Nnt siRNAs. Each siRNA had a Cy3 tag Protein content was measured the DC protein assay kit (BioRad, Hercules, on the 50 end of the sense strand to enable visualization of transfected cells. California). ATP content was normalized to protein content to correct for dif- siRNA duplexes (5 nM) were transfected into MIN6 cells using Lipofectamine ferences in islet size. 2000 (Invitrogen) at 200 nM. Cells were cultured in DMEM (FBS-free) for a further 24 hr before experiment. Glucose utilization measurements 3 Quantitative RT-PCR Glucose utilization was determined as the formation of [ H]water from 3 Total RNA from siRNA-transfected MIN6 cells was extracted using RNeasy [5- H]glucose. After overnight culture in PIM medium, batches of five islets 3 mini-kit (Qiagen). cDNA generated by Superscript II enzyme (Invitrogen) were incubated at 37ºCfor2hrin10ml of KRB solution containing [5- H]glu- was analyzed by RT-PCR using a SYBR green PCR kit and an ABIPRISM cose at various concentrations of glucose, as described by (Ashcroft et al., 7700 Sequence detector (Perkin Elmer). To quantitate Nnt expression, it 1972). Control samples containing no islets were included to allow for correc- 3 3 was PCR amplified from cDNA together with Gapdh (glyceraldehyde-3- tion for [ H]water in [ H]glucose. Glucose utilization rates were calculated as 3 phosphate dehydrogenase) as an endogenous control. All data were normal- pmol of glucose utilized/hour/islet from: Glucose utilized (pmol) = [ H]water 3 ized to Gapdh expression. RT-PCR primers listed in 50 to 30 orientation are: formed (c.p.m)/specific radioactivity of [5- H]glucose (c.p.m./pmol). For Gapdh: Gapdh888F: CCTGCGACTTCAACAGCAACT and Gapdh979RV: CCAGGAAATGAGCTTGACAAA; For Nnt: NntXn5inF: GATAGTTGGTGGTG Measurement of reactive oxygen species 0 0 GAGTTG and NntXn6inR: GATGTCCACCTCCTTGCACT. A 10 mM stock solution of H2DCFDA-SE (2 ,7 -dichlorofluorescein diacetate The t tests for significant differences between strains are based on RQ with a succinimidyl ester group, Molecular Probes) was prepared in DMSO, (Nnt/Gapdh) values. stored at 270ºC, and diluted with KRB (pH 7.0) just before use to a final concentration of 10 mM. b cells were plated in 24-well cell culture plates over- Protein gels and WesternRETRACTED blot analysis night. They were then washed twice with PBS, incubated in PBS plus DCF Target-specific gene knockdown was verified using SDS-PAGE 12% gel and for 30 min at 37ºC, washed with PBS, and incubated in KRB containing Western blot. Protein samples were heated for 3 min at 100ºC in sample 0 or 20 mM glucose for a further 30 min. They were imaged at room temper- buffer (final concentration of 30 mM Tris-HCl [pH 6.8], 5% glycerol, ature using an IonOptix fluorescence system (Boston, Massachusetts), with 0.005% bromophenol blue, 2.5% b-mercaptoethanol). Proteins were size- 495 nm excitation and 520 nm emission. The background fluorescence was separated by electrophoresis alongside a ‘‘SeeBlue’’ (Invitrogen) protein lad- subtracted. We used 0.5 mM menadione as a generator of ROS as a control: der to allow molecular weights to be estimated. we observed no difference in fluorescence between wild-type and Nnt Proteins were transferred to Hybond-P membrane (Amersham) using mouse b cells exposed to menadione, indicating that there were no differen- a Transblot SD system (Biorad) by passing 100V across membrane/gel for ces in dye loading between wild-type and mutant b cells.

CELL METABOLISM : JANUARY 2006 43 ARTICLE

Supplemental data lecular defects of pancreatic islets in human type 2 diabetes. Diabetes 54, 727–735. Supplemental Data include three ﬁgures and Supplemental Experimen- tal Procedures and can be found with this article online at http://www. Echtay, K.S., Roussel, D., St-Pierre, J., Jekabsons, M.B., Cadenas, S., Stu- cellmetabolism.org/cgi/content/full/3/1/35/DC1/. art, J.A., Harper, J.A., Roebuck, S.J., Morrison, A., Pickering, S., et al. (2002). Superoxide activates mitochondrial uncoupling proteins. Nature 415, 96–99.

Acknowledgments Fridlyand, L.E., and Philipson, L.H. (2004). Does the glucose-dependent insulin secretion mechanism itself cause oxidative stress in pancreatic beta- We thank the Medical Research Council (to R.D.C. and Studentship to H.F.), cells? Diabetes 53, 1942–1948. the Wellcome Trust (F.M.A.), Diabetes UK (grant RD00/0002072 to R.D.C.), and the Royal Society (F.M.A.) for support. We thank Dr Jon Lippiat for Friesen, N.T., Buchau, A.S., Schott-Ohly, P., Lgssiar, A., and Gleichmann, H. help with early experiments using siRNA and Drs Martin Brand and Karl (2004). Generation of hydrogen peroxide and failure of antioxidative responses Morten for advice. in pancreatic islets of male C57BL/6 mice are associated with diabetes induced by multiple low doses of streptozotocin. Diabetologia 47, 676–685.

Gloyn, A.L., Pearson, E.R., Antcliff, J.F., Proks, P., Bruining, G.J., Slingerland, A.S., Howard, N., Srinivasan, S., Silva, J.M., Molnes, J., et al. (2004). Activat- Received: June 21, 2005 ing mutations in the gene encoding the ATP-sensitive potassium-channel Revised: September 27, 2005 subunit Kir6.2 and permanent neonatal diabetes. N. Engl. J. Med. 350, Accepted: October 21, 2005 1838–1849. Published: January 10, 2006 Green, K., Brand, M.D., and Murphy, M.P. (2004). Prevention of mitochon- References drial oxidative damage as a therapeutic strategy in diabetes. Diabetes 53 (Suppl 1), S110–S118. Arkblad, E.L., Betsholtz, C., Mandoli, D., and Rydstrom, J. (2001). Character- Hoek, J.B., and Rydstrom, J. (1988). Physiological roles of nicotinamide nu- ization of a nicotinamide nucleotide transhydrogenase gene from the green cleotide transhydrogenase. Biochem. J. 254, 1–10. alga Acetabularia acetabulum and comparison of its structure with those of the corresponding genes in mouse and Caenorhabditis elegans. Biochim. Kahn, S.E. (2003). The relative contributions of insulin resistance and beta- Biophys. Acta 1520, 115–123. cell dysfunction to the pathophysiology of Type 2 diabetes. Diabetologia Arkblad, E.L., Egorov, M., Shakhparonov, M., Romanova, L., Polzikov, M., 46, 3–19. and Rydstrom, J. (2002). Expression of proton-pumping nicotinamide nucle- Kaku, K., Fiedorek, F.T., Jr., Province, M., and Permutt, M.A. (1988). Genetic otide transhydrogenase in mouse, human brain and C elegans. Comp. Bio- analysis of glucose tolerance in inbred mouse strains. Evidence for polygenic chem. Physiol. B Biochem. Mol. Biol. 133, 13–21. control. Diabetes 37, 707–713. Arkblad, E.L., Tuck, S., Pestov, N.B., Dmitriev, R.I., Kostina, M.B., Stenvall, Kayo, T., Fujita, H., Nozaki, J., E, X., and Koizumi, A. (2000). Identiﬁcation of J., Tranberg, M., and Rydstrom, J. (2005). A Caenorhabditis elegans mutant two chromosomal loci determining glucose intolerance in a C57BL/6 mouse lacking functional nicotinamide nucleotide transhydrogenase displays in- strain. Comp. Med. 50, 296–302. creased sensitivity to oxidative stress. Free Radic. Biol. Med. 38, 1518–1525.

Ashcroft, F., and Rorsman, P. (2004). Type 2 diabetes mellitus: not quite ex- King, H., Aubert, R.E., and Herman, W.H. (1998). Global burden of diabetes, citing enough? Hum. Mol. Genet. 13, R21–R31. 1995–2025: prevalence, numerical estimates, and projections. Diabetes Care 21, 1414–1431. Ashcroft, F.M., and Kakei, M. (1989). ATP-sensitive K+ channels in rat pancreatic beta-cells: modulation by ATP and Mg2+ ions. J. Physiol. 416, 349– Kooptiwut, S., Zraika, S., Thorburn, A.W., Dunlop, M.E., Darwiche, R., Kay, 367. T.W., Proietto, J., and Andrikopoulos, S. (2002). Comparison of insulin secretory function in two mouse models with different susceptibility to beta-cell Ashcroft, F.M., and Rorsman, P. (1989). Electrophysiology of the pancreatic failure. Endocrinology 143, 2085–2092. beta-cell. Prog. Biophys. Mol. Biol. 54, 87–143. Koster, J.C., Marshall, B.A., Ensor, N., Corbett, J.A., and Nichols, C.G. Ashcroft, S.J., Weerasinghe, L.C., Bassett, J.M., and Randle, P.J. (1972). The (2000). Targeted overactivity of beta cell K(ATP) channels induces profound pentose cycle and insulin release in mouse pancreatic islets. Biochem. J. neonatal diabetes. Cell 100, 645–654. 126, 525–532. Krauss, S., Zhang, C.Y., Scorrano, L., Dalgaard, L.T., St-Pierre, J., Grey, S.T., Bell, G.I., and Polonsky, K.S. (2001). Diabetes mellitus and genetically pro- and Lowell, B.B. (2003). Superoxide-mediated activation of uncoupling pro- grammed defects in beta-cell function. Nature 414, 788–791. tein 2 causes pancreatic beta cell dysfunction. J. Clin. Invest. 112, 1831– Bindokas, V.P., Kuznetsov, A., Sreenan, S., Polonsky, K.S., Roe, M.W., and 1842. Philipson, L.H. (2003). Visualizing superoxide production in normal and dia- Krippeit-Drews, P., Kramer, C., Welker, S., Lang, F., Ammon, H.P., and betic rat islets of Langerhans. J. Biol. Chem. 278, 9796–9801. Drews, G. (1999). Interference of H2O2 with stimulus-secretion coupling in Brand, M.D., Affourtit, C., Esteves, T.C., Green, K., Lambert, A.J., Miwa, S., mouse pancreatic beta-cells. J. Physiol. 514, 471–481. Pakay, J.L., and Parker, N. (2004). Mitochondrial superoxide: production, bi- ological effects, and activation of uncoupling proteins. Free Radic. Biol. Med. Kulkarni, R.N., Almind, K., Goren, H.J., Winnay, J.N., Ueki, K., Okada, T., and 37, 755–767. Kahn, C.R. (2003). Impact of genetic background on development of hyper- insulinemia and diabetes in insulin receptor/insulin receptor substrate-1 dou- Chan, C.B., De Leo, D., Joseph, J.W., McQuaid, T.S., Ha, X.F., Xu, F., Tsush- ble heterozygous mice. Diabetes 52, 1528–1534. ima, R.G., Pennefather, P.S., Salapatek, A.M., and Wheeler, M.B. (2001). In- RETRACTEDMaechler, P., Jornot, L., and Wollheim, C.B. (1999). Hydrogen peroxide alters creased uncoupling protein-2 levels in beta-cells are associated with impaired glucose-stimulated insulin secretion: mechanism of action. Diabetes mitochondrial activation and insulin secretion in pancreatic beta cells. J. Biol. 50, 1302–1310. Chem. 274, 27905–27913.

Coghill, E.L., Hugill, A., Parkinson, N., Davison, C., Glenister, P., Clements, Maechler, P., and Wollheim, C.B. (2001). Mitochondrial function in normal S., Hunter, J., Cox, R.D., and Brown, S.D. (2002). A gene-driven approach and diabetic beta-cells. Nature 414, 807–812. to the identiﬁcation of ENU mutants in the mouse. Nat. Genet. 30, 255–256. Martens, G.A., Cai, Y., Hinke, S., Stange, G., Van de Casteele, M., and Pipel- Del Guerra, S., Lupi, R., Marselli, L., Masini, M., Bugliani, M., Sbrana, S., eers, D. (2005). Glucose suppresses superoxide generation in metabolically Torri, S., Pollera, M., Boggi, U., Mosca, F., et al. (2005). Functional and mo- responsive pancreatic beta cells. J. Biol. Chem. 280, 20389–20396.

44 CELL METABOLISM : JANUARY 2006 Nnt and insulin secretion

Muller, F.L., Liu, Y., and Van Remmen, H. (2004). Complex III releases super- Silva, J.P., Kohler, M., Graff, C., Oldfors, A., Magnuson, M.A., Berggren, oxide to both sides of the inner mitochondrial membrane. J. Biol. Chem. 279, P.O., and Larsson, N.G. (2000). Impaired insulin secretion and beta-cell 49064–49073. loss in tissue-speciﬁc knockout mice with mitochondrial diabetes. Nat. Genet. 26, 336–340. Proks, P., Antcliff, J.F., Lippiat, J., Gloyn, A.L., Hattersley, A.T., and Ashcroft, F.M. (2004). Molecular basis of Kir6.2 mutations associated with neonatal di- Stoehr, J.P., Nadler, S.T., Schueler, K.L., Rabaglia, M.E., Yandell, B.S., Metz, abetes or neonatal diabetes plus neurological features. Proc. Natl. Acad. Sci. S.A., and Attie, A.D. (2000). Genetic obesity unmasks nonlinear interactions USA 101, 17539–17544. between murine type 2 diabetes susceptibility loci. Diabetes 49, 1946–1954.

Quwailid, M.M., Hugill, A., Dear, N., Vizor, L., Wells, S., Horner, E., Fuller, S., Terauchi, Y., Sakura, H., Yasuda, K., Iwamoto, K., Takahashi, N., Ito, K., Ka- Weedon, J., McMath, H., Woodman, P., et al. (2004). A gene-driven ENU- sai, H., Suzuki, H., Ueda, O., Kamada, N., et al. (1995). Pancreatic beta-cell- based approach to generating an allelic series in any gene. Mamm. Genome speciﬁc targeted disruption of glucokinase gene. Diabetes mellitus due to 15, 585–591. defective insulin secretion to glucose. J. Biol. Chem. 270, 30253–30256. Toye, A.A., Lippiat, J.D., Proks, P., Shimomura, K., Bentley, L., Hugill, A., Mi- Raha, S., and Robinson, B.H. (2000). Mitochondria, oxygen free radicals, dis- jat, V., Goldsworthy, M., Moir, L., Haynes, A., et al. (2005). A genetic and ease and ageing. Trends Biochem. Sci. 25, 502–508. physiological study of impaired glucose homeostasis control in C57BL/6J mice. Diabetologia. Rossmeisl, M., Rim, J.S., Koza, R.A., and Kozak, L.P. (2003). Variation in type 2 diabetes–related traits in mouse strains susceptible to diet-induced obe- Tucker, S.J., Gribble, F.M., Zhao, C., Trapp, S., and Ashcroft, F.M. (1997). sity. Diabetes 52, 1958–1966. Truncation of Kir6.2 produces ATP-sensitive K+ channels in the absence of the sulphonylurea receptor. Nature 387, 179–183. Sakura, H., Ammala, C., Smith, P.A., Gribble, F.M., and Ashcroft, F.M. (1995). Cloning and functional expression of the cDNA encoding a novel ATP-sensi- Yamaguchi, M., and Hateﬁ, Y. (1991). Mitochondrial energy-linked nicotin- tive potassium channel subunit expressed in pancreatic beta-cells, brain, amide nucleotide transhydrogenase. Membrane topography of the bovine heart and skeletal muscle. FEBS Lett. 377, 338–344. enzyme. J. Biol. Chem. 266, 5728–5735.

Sazanov, L.A., and Jackson, J.B. (1994). Proton-translocating transhydroge- Zhang, C.Y., Baffy, G., Perret, P., Krauss, S., Peroni, O., Grujic, D., Hagen, T., nase and NAD- and NADP-linked isocitrate dehydrogenases operate in Vidal-Puig, A.J., Boss, O., Kim, Y.B., et al. (2001). Uncoupling protein-2 neg- a substrate cycle which contributes to ﬁne regulation of the tricarboxylic atively regulates insulin secretion and is a major link between obesity, beta acid cycle activity in mitochondria. FEBS Lett. 344, 109–116. cell dysfunction, and type 2 diabetes. Cell 105, 745–755.

RETRACTED

CELL METABOLISM : JANUARY 2006 45