Sirt1 Regulates Insulin Secretion by Repressing UCP2 in Pancreatic β Cells

Sir2 and insulin/IGF-1 are the major pathways that impinge upon aging in lower organisms. In Caenorhabditis elegans a possible genetic link between Sir2 and the insulin/IGF-1 pathway has been reported. Here we investigate such a link in mammals. We show that Sirt1 positively regulates insulin secretion in pancreatic β cells. Sirt1 represses the uncoupling protein (UCP) gene UCP2 by binding directly to the UCP2 promoter. In β cell lines in which Sirt1 is reduced by SiRNA, UCP2 levels are elevated and insulin secretion is blunted. The up-regulation of UCP2 is associated with a failure of cells to increase ATP levels after glucose stimulation. Knockdown of UCP2 restores the ability to secrete insulin in cells with reduced Sirt1, showing that UCP2 causes the defect in glucose-stimulated insulin secretion. Food deprivation induces UCP2 in mouse pancreas, which may occur via a reduction in NAD (a derivative of niacin) levels in the pancreas and down-regulation of Sirt1. Sirt1 knockout mice display constitutively high UCP2 expression. Our findings show that Sirt1 regulates UCP2 in β cells to affect insulin secretion.


Introduction
Glucose homeostasis is maintained, in part, by pancreatic b cells, which secrete insulin in a highly regulated sequence of dependent events [1]. b cells metabolize glucose, resulting in an increase in the ATP/ADP ratio, the closing of the ATPdependent K þ channel, the activation of the voltage-gated Ca þ channel and Ca þ influx, and the fusion of secretory vesicles to the plasma membrane to release insulin. Insulin is part of an organismal physiological axis in which it stimulates glucose uptake in metabolic tissues, such as muscle, and stores energy in the form of fat in white adipose tissue (WAT). Short-term food limitation (i.e., overnight [O/N] fasting) will therefore elicit the mobilization of glycogen stores and then fat from WAT for metabolism, and the lower level of blood glucose during fasting will result in low levels of insulin production by b cells.
Long-term calorie restriction (CR) has been known for 70 years to extend the life span of mammals dramatically [2], and it can also work in a variety of organisms, including yeast, flies, and rodents [3][4][5], although the mechanism of this effect has remained obscure. In mammals a characteristic set of physiological changes takes place during long-term CR, which overlaps the rapid physiological adaptations to short-term food limitation. One such change is the use of dietary fat or fat mobilized from WAT for energy [4]. Another is a large reduction in blood insulin levels accompanied by an increase in insulin sensitivity, i.e., the ability of insulin to promote glucose utilization [4]. In addition, gluconeogenesis is activated in the liver. These changes keep glucose available for the brain, and are closely associated with the longevity elicited by CR. The paucity of fat in WAT appears to be sufficient per se to promote a degree of longevity, since mice engineered for leanness-for example, a WAT-specific knockout (KO) of the insulin receptor-live longer [6,7].
Findings in model organisms suggest a mechanism for the longevity engendered by CR that implicates the silent mating type information regulation 2 gene (Sir2). This gene regulates the life span in yeast [8] and Caenorhabditis elegans [9] as a longevity determinant. In yeast, CR works by up-regulating the activity of Sir2 [10,11], a NAD-dependent deacetylase [12][13][14] (NAD is a derivative of niacin), by increasing respiration, and by increasing the NAD/NADH ratio [15] (NADH is the reduced form of NAD). CR is also reported to activate the NAD salvage pathway, which would deplete a Sir2 inhibitor, nicotinamide [3,10]. The Drosophila melanogaster Sir2 gene was also shown to mediate life extension in response to dietary restriction [16,17].
Since Sir2 appears to mediate the effects of CR on life span in simple model organisms, it seemed possible that Sir2 proteins also regulate the effects of food limitation and CR in mammals. The homolog of the yeast silencing information regulator2 (Sirt1) has also been implicated in several aspects of food limitation and CR in mammals. In WAT, Sirt1 represses the key regulatory protein peroxisome proliferatoractivated receptor gamma (PPARc), resulting in fat mobilization in response to food limitation [18]. In addition, Sirt1 regulates the FOXO (forkhead Box O) set of forkhead transcription factors [19,20], providing another link to metabolism and diet. Also, gluconeogenesis in the liver is regulated by Sirt1 [19], which works in concert with the transcriptional co-activator, peroxisome proliferator-activated receptor coactivator, PGC-1a [21]. Finally, Sirt1 may play a role in the observed stress resistance of CR animals, since it down-regulates several pro-apoptotic factors, such as p53, FOXO, and Bax [19,20,[22][23][24][25].
In addition to the classical paradigm for insulin regulation by glucose outlined above, reports suggest a role of an uncoupling protein (UCP) in insulin secretion. UCPs belong to a family of mitochondrial inner membrane proteins. They function to uncouple oxygen consumption during respiration from the production of ATP by allowing proton leakage down an electrochemical gradient from the cytoplasm to the mitochondria [26][27][28][29]. Several mammalian UCP homologues have been identified and characterized [28]. UCP2 has been shown to promote proton leakage across the mitochondrial membrane [30][31][32][33], and has been proposed to play a role in lipid metabolism, insulin resistance, glucose utilization, regulation of reactive oxygen species, and macrophagemediated immunity [28,34,35]. A role for UCP2 in insulin secretion has been demonstrated, since UCP2 KO mice show higher ATP levels in islets and increased insulin secretion upon glucose stimulation [36]. Conversely, UCP2 overexpression in cultured b cell lines reduces ATP levels and glucosestimulated insulin secretion [37,38].
In this study, we show that Sirt1 functions as a positive regulator of insulin secretion in response to glucose by directly repressing the UCP2 gene. Our findings suggest that UCP2 is thus used in b cells to modulate the insulin response pathway as a function of diet. The fact that Sirt1 is a positive regulator of insulin may seem surprising, in light of the fact that sir-2.1 in C. elegans appears to be a negative regulator of the insulin-like signaling pathway [9]. We discuss how this difference makes sense in light of the physiological adaptations that take place in mammals in response to dietary changes.

Sirt1 Is Preferentially Expressed in Pancreatic b Cells
To determine whether Sirt1 could potentially play a role in b cells, immunofluorescence of whole murine pancreas with anti-Sirt1 antibody was carried out, and 20 islets were examined by fluorescence microscopy. A representative islet The pancreas of wild-type mice was sectioned and stained as described.
(A) Nuclear staining using DAPI (top left). Immunofluorescence using Sirt1 antibody (top right); hematoxylin and eosin staining of the same section of pancreas (bottom left); immunofluorescence control using a rabbit secondary antibody (bottom right). (B) Pancreases of wild-type (WT), Sirt1þ/À heterozygotes (HET), or Sirt1À/À homozygous KO mice were stained with antibodies against insulin (blue), glucagon (red), or somatostatin (green) (shown in left column). Representative islets of mice of all three genotypes are shown. Pancreases were also silver-stained for morphometry (right column). Islets appear as dark figures and their area was determined by scanning, using Image-Pro 4. is shown in Figure 1A (see lower left panel for hematoxylineosin staining). All pancreatic sections examined showed intense staining concentrated in islets using Sirt1 antibodies ( Figure 1A, top right) but not with secondary antibody alone ( Figure 1A, lower right), and a lower level of expression in the surrounding exocrine cells. DAPI bright spots ( Figure 1A, top left) mark nuclei and their corresponding cells in the islets and surrounding tissue. This enrichment of Sirt1 in the islets and not exocrine cells is noteworthy, in light of the ubiquitous expression of this sirtuin in most somatic and germ tissues [39,40].

Sirt1 KO Mice Have Low Levels of Blood Insulin
Because Sirt1 is highly expressed in b cells, we investigated whether this sirtuin plays a functional role in insulin production. First, we determined whether Sirt1 KO mice [39,40] showed any defects in the pancreatic b cells or the islets. Pancreases from wild-type, Sirt1þ/À heterozygotes, or Sirt1À/À KO mice (4À5 each) were sectioned and stained with antibodies against insulin (blue), glucagons (red), and somatostatin (green). These stains mark islets for b cells, a cells, and d cells, respectively. As shown in a typical section ( Figure 1B), no differences were observed in the staining pattern of these three markers between wild-type and mutant islets. We measured pancreatic islet areas using Image-Pro 4.1 Plus software, and expressed the islet areas as a percentage of the total pancreatic area ( Figure 1C). There were no significant differences in islet areas comparing wild-type, heterozygous, and KO mice. The absolute islet size was also not appreciably different in the mutant mice.
Next, we determined whether insulin production was altered in Sirt1 KO mice. Blood insulin was measured in males 2-4 mo old fed ad libitum or after 18 h of starvation ( Figure 2A). Sirt1 KO mice (black bars) had much lower blood levels of insulin compared with littermate control animals (open bars) when the samples were collected under ad libitum conditions. This difference was also observed when the animals were starved O/N, in which case insulin levels were very low in both wild-type and KO mice. To assess more precisely the ability of animals to produce insulin, mice were given an injection of glucose after O/N starvation and insulin was measured after 2, 10, and 20 min. Whereas insulin induction was clearly evident in the wild-type mice at 10 min (open bars), induction was not observed in Sirt1 KO animals (black bars in Figure 2B, n ¼ 4 or 5 for each measurement). A similar trend was noted at 20 min.
To further investigate the defect in insulin production, islets from four wild-type or four Sirt1 KO animals were isolated and incubated in vitro with or without glucose for 1 h, and insulin secretion was determined. As shown in Figure  2C, the basal level of insulin secreted into the media from islets of Sirt1 KO mice (black bars) was significantly lower than wild-type controls (open bars). Moreover, the islets isolated from Sirt1 KO mice were not induced to secrete insulin by glucose, while control islets were inducible, as expected.
Levels of blood glucose were then determined and, surprisingly, were lower in Sirt1 KO mice ( Figure 2D). Further, these mice appeared to be hypermetabolic, since they ate more food per body weight than the wild-type (unpublished data). These findings suggested that the KO mice had a better ability to use the lower levels of insulin for glucose uptake, i.e., were more insulin sensitive. To address this possibility, we performed glucose tolerance tests in both wild-type and KO mice by injecting glucose intraperitoneally and measuring the kinetics of glucose clearance from the blood. The KO mice cleared the glucose significantly faster than the wild-type ( Figure 2E). This increased glucose tolerance explains why the KO mice maintained lower levels of blood glucose, even with reduced levels of insulin. The surprising effect of knocking out Sirt1 on glucose tolerance likely derives from tissues other than the pancreas and is currently being investigated in greater detail. In summary, we do not know whether the reduction in insulin in Sirt1 KO mice is due to a b cell defect or is an indirect consequence of increased glucose tolerance in these animals.

Sirt1 Drives Glucose-Induced Insulin Secretion in Cultured Cells
To address whether Sirt1 played a positive role in insulin production specifically in b-cells, the pancreatic b-cell lines, INS-1 from rat, and MIN6 from mouse were employed. INS-1 cells were grown in culture, and nuclear expression of Sirt1 was evident by immunofluorescence ( Figure 3) and Western blot ( Figure 3C and 3E). INS-1 or MIN6 cells were treated with 10 mM nicotinamide, an inhibitor of Sirt1 [41], for 48 h, and the levels of insulin secreted into the media were measured without and with induction by glucose ( Figure 3B and 3F, respectively). Whereas control cells showed robust induction of insulin secretion by glucose, nicotinamide treatment blocked this induction in both cell lines.
To test if the effects of nicotinamide were due specifically to inhibition of Sirt1, levels of this sirtuin were lowered by RNA interference. INS-1 and MIN6 cells were infected with a retrovirus carrying a Sirt1-SiRNA construct [18] or a control vector (pSUPER-SiRNA GFP), and stable cell lines were created as puromycin-resistant infected pools. These cells displayed Sirt1 protein levels that were knocked down compared with drug-resistant control pools generated from the vector ( Figure 3C [INS-1] and 3E [MIN]). When these stable Sirt1 knockdown cells ( Figure 3D, 3F, and 3G) were treated with glucose, induction of insulin secretion was eliminated, whereas contemporaneous assays of vector control cells (open bars) showed normal induction ( Figure 3D and 3F).
We then tested if the defect in glucose-induced insulin secretion was due to a different glucose uptake between control and knockdown cells by measuring transport of fluorescence analog 2-NBDG. Glucose uptake [42,43] was not reduced and was perhaps slightly elevated in this assay in the knockdown cells ( Figure 3G). Further, no decrease in the glucose transporter Glut 2 was observed (unpublished data). Finally, cell growth measurements revealed no difference in the growth rate of INS-1 knockdown cells compared with controls ( Figure S1). These results show that knockdown of Sirt1 suppresses glucose-stimulated insulin secretion in two different b cell lines.

ATP/ADP Ratio Is Lower in Sirt1 Knockdown Cells
We next sought to investigate the mechanism by which Sirt1 regulates insulin secretion in b cells. First, we assayed the RNA level of the INS1 gene, encoding insulin, by Northern blot and found no difference in control versus Sirt1 knockdown b cells ( Figure 4A). Similarly, we found no difference by Western blot in expression level of c/EBPb, a transcription factor that regulates INS1 ( Figure 4B). Because the insulin receptor is part of an autocrine induction loop for insulin [44], we also determined by Western blot the levels of this receptor (a and b subunits) and observed no difference between control and knockdown cells ( Figure 4B). Next, we investigated whether Sirt1 affected the levels of the K þ channel by Western blot, and again found no effect ( Figure  4C). While this analysis suggests that Sirt1 does not function by altering levels of these factors, it does not rule out the possibility that Sirt1 regulates their activity.
Finally, because of the central role of ATP in insulin secretion, we surmised that Sirt1 could play a role in the energetics of glucose utilization in b cells. We thus measured the ATP/ADP ratio in control and Sirt1 knockdown INS-1 cells after glucose induction. Control cells ( Figure 4D, open bars) responded to glucose by increasing the ratio of ATP to ADP, as expected ( Figure 4D). In contrast, in cells with the knockdown levels of Sirt1 ( Figure 4D, black bars), the ATP/ ADP ratio did not increase upon glucose stimulation. These findings show that knocking down Sirt1 results in a defect in ATP production in response to glucose. Basal ATP levels are not significantly altered in knockdown cells, consistent with the fact that they grow as well as control cells.

UCP2 Levels Are Increased in Sirt1 KO Mice and Knockdown Cells
One possibility for the failure of the Sirt1 knockdown cells to make ATP is that respiration is more uncoupled than in control cells, which would square well with the known link  between the UCP2 and insulin production in b cells (36). Thus, we determined by Western blot the levels of UCP2 in control and Sirt1 knockdown cells. Strikingly, there was a significant increase in the UCP2 protein level in the knockdown cells ( Figure 4E). This increase in protein level was mirrored by an increase in the mRNA in the same cells ( Figure 4F), indicating that Sirt1 regulates UCP2 transcription. It was previously shown that UCP2 expression reduced NADH levels [45]. We therefore determined NADH levels by autofluorescence [46] and found a significantly lower level of NADH in the knockdown cells ( Figure 4G). Western blot for UCP2 in Sirt1 KO mice showed a similar effect. We observed an increase in UCP2 protein in the whole pancreas (unpublished data), which is a measure of the islets, since UCP2 is expressed in only the endocrine cells of the pancreas [38]. Moreover, UCP2 protein was also up-regulated in isolated islets of Sirt1 KO mice ( Figure 4H). In summary, our expression studies suggest that Sirt1 is a repressor of UCP2 transcription in b cells, and by repressing this UCP, this sirtuin may allow cells to secrete insulin in response to glucose.

Sirt1 Binds to the UCP2 Promoter
To study further the repression of UCP2 by Sirt1, we carried out reporter assays in 293T cells transfected with a chloramphenicol acetyl-transferase (CAT) gene, whose expression is driven by the UCP2 promoter. Cells were also transfected with a control vector or with an expression vector for PPARc, which is known to bind to and activate the UCP2 promoter [47]. In a control experiment, PPARc activated the reporter in this assay ( Figure 5A, left bars). In a parallel experiment, cells were also co-transfected with a Sirt1 expression vector. Sirt1 clearly repressed the activation of the UCP2 promoter ( Figure 5A, right bars). Repression of this reporter by endogenous or expressed Sirt1 was alleviated by nicotinamide, a known inhibitor of Sirt1 (unpublished data).
To determine whether repression of UCP2 was due to the direct binding of Sirt1 at the promoter, INS-1 cells were subjected to chromatin-immunoprecipitation, using Sirt1 or control antibodies. Primers specifically designed to span a known regulatory region of the UCP2 promoter (43) ( Figure  5B, primer set 5) were used to probe by PCR the DNA in the immunoprecipitate. Sirt1 bound to this region of the UCP2 promoter in the control cells ( Figure 5C, column 3) and to a significantly lesser extent in the Sirt1 knockdown cells ( Figure  5C, column 6). Control primers designed in a different region of UCP2 upstream DNA ( Figure 5B, primer set 4) showed no amplification ( Figure 5C, columns 2 and 5). An anti-gal4 antibody control using primer set 5 also showed no binding ( Figure 5C, columns 1 and 4). These findings indicate that Sirt1 represses UCP2 transcription by binding directly at the UCP2 promoter.

Reduction of UCP2 Restores Insulin Secretion in Sirt1 Knockdown b cells
The above findings show that Sirt1 represses UCP2, and alleviation of this repression correlated with blunted insulin secretion in response to glucose. To address whether the increase in UCP2 in Sirt1 knockdown cells caused the failure to secrete insulin, UCP2 was also knocked down in INS-1 cells with reduced Sirt1. Stable cell lines with UCP2 knocked down by SiRNA were derived from two different lines in which Sirt1 was already knocked down, as well as from control INS-1 cells. Several different UCP2 sequences were inserted into a pSUPER hairpin vector with the neomycin-resistance drug marker. These constructs were transfected into the SiRNA Sirt1 puromycin resistant cells or control cells, and stable populations of NeoR cells were assayed for UCP2 RNA and protein by Northern and Western blots. In the case of two different UCP2 SiRNA constructs, the levels of UCP2 RNA ( Figure 6A) and protein (unpublished data) were clearly reduced in transfected cells. The levels of Sirt1 remained low in the double knockdown cells (unpublished data).
Next, glucose-stimulated insulin secretion was assayed in Sirt1 or UCP2 knockdown cells and in cells with both Sirt1 and UCP2 knocked down. The Sirt1 knockdown cells were again defective in induction of insulin secretion, as expected. However, the double knockdown cells or cells with only the UCP2 SiRNA construct displayed insulin secretion in response to glucose ( Figure 6B). A similar effect was observed in the second Sirt1 knockdown line in which UCP2 was also knocked down (unpublished data). This result demonstrates that the failure of Sirt1 knockdown cells to secrete insulin is due to the elevated levels of UCP2 in these cells. We conclude that Sirt1 acts as a positive regulator of insulin secretion in wild-type cells by repressing UCP2, thereby allowing coupling of glucose metabolism to ATP synthesis.

UCP2 Levels Increase in Food-Deprived Mice
Does the regulation of UCP2 play any role in the normal secretion of insulin in b cells of wild-type mice in response to diet? To address this question, we starved wild-type mice O/N and compared the levels of UCP2 in whole pancreas and in islets to mice feeding ad libitum. Importantly, we found an increase in UCP2 mRNA levels in whole pancreas of starved mice compared with the fed mice ( Figure 7A). Further, the levels of UCP2 protein were also increased in starved mice ( Figure 7B).
In order to determine whether Sirt1 regulated this induction of UCP2 in mice, we starved Sirt1 KO mice O/N and compared the effect of starvation on UCP2 protein levels to wild-type mice. Four pairs of wild-type and KO littermates were compared and gave comparable results. The levels of UCP2 in KO mice fed ad libitum were elevated compared with wild-type, as expected ( Figure 7C). Most importantly, these elevated levels in the fed KO mice were not further induced by starvation. In contrast, starvation induced UCP2 in wildtype mice, as before. A similar pattern of UCP2 RNA induction by starvation in wild-type but not KO mice was observed by RT-PCR ( Figure 7D). The above findings suggest that an increase in UCP2 in b cells is part of a normal mechanism to regulate the capacity of b cells to produce insulin. Moreover, this induction appears to be mediated by alleviation of Sirt1-mediated repression.
These findings suggest that starvation causes a decrease in Sirt1 activity in b cells. In yeast, Sir2p activity is regulated by the NAD/NADH ratio. We thus measured NAD and NADH in the pancreas of seven fed and seven starved wild-type mice. Strikingly, there was a significant decrease in the level of NAD but not NADH in the starved mice ( Figure 8). The level of Sirt1 protein in starved pancreas is roughly comparable to fed controls (unpublished data). These findings suggest that changes in NAD levels in the pancreas regulate Sirt1 activity and insulin secretion in response to diet.

Discussion
In this study we show that Sirt1 regulates insulin secretion in pancreatic b cells. Both in Sirt1 KO mice and in cultured b cell lines, a reduction in Sirt1 levels reduces the capacity of these cells to secrete insulin in response to glucose. Classical  studies show that an increase in the ATP/ADP ratio due to glucose metabolism is the critical trigger in the induction of insulin secretion. In cells with reduced Sirt1, the increase in this ratio is blunted due to elevated levels of UCP2, which reduces the synthesis of ATP during respiration. Sirt1 directly represses UCP2 by binding to the promoter. Thus, by controlling UCP2, Sirt1 regulates the amplitude of insulin induction by glucose. We do not know whether the low levels of insulin in the whole body of Sirt1 KO mice represents a b cell defect, an indirect consequence of the elevated glucose tolerance in these animals, or both.
The use of UCP2 to regulate insulin production is distinct from the canonical role of UCPs. By uncoupling respiration from ATP synthesis in brown fat, for example, UCP1 generates heat and is critical for the non-shivering thermogenesis in rodents (reviewed in [48]). Here UCP2 is used to modulate the levels of ATP made in b cells, which dictates the extent of insulin secretion. Our findings are consistent with the elevated basal levels of insulin observed in fasted UCP2 KO mice [36] and suggest that Sirt1 and UCP2 are physiologically relevant regulators of insulin production (see ''UCPs and Longevity-Cellular Basis'').

Physiology of Insulin Regulation by Sirt1
The fact that Sirt1 represses UCP2 expression in b cells is intriguing in light of the role of this sirtuin as a regulator of CR in lower organisms. Our findings indicate that Sirt1 may link the amplitude of insulin secretion to the diet. We found that O/N starvation led to an increase in the levels of UCP2 protein and RNA in the pancreas and isolated islets. This increase did not occur in Sirt1 KO mice, which already displayed higher levels of UCP2 when fed ad libitum. These findings suggest that regulation of UCP2 by Sirt1 is a physiologically important mechanism to blunt the chronic levels of insulin in starved animals. This mechanism may explain why insulin levels are so low during fasting, even though b cells are actively metabolizing fat and respiring (see following section). It remains to be seen whether this mechanism is important during long-term CR.

Regulation of Pancreatic Sirt1 Activity by NAD/NADH and Diet
The action of mammalian Sirt1 appears to differ from that of lower organisms. In C. elegans, sir2.1 appears to repress the output of the insulin/IGF pathway [9], but the mechanism has not been described. Also, Sir2 activity is activated during CR to extend the life span in yeast and Drosophila [11,49]. However, in mammals Sirt1 functions as a positive regulator of insulin secretion, raising the possibility that its activity in b cells is actually reduced by O/N starvation..
A lowering of Sirt1 activity occurs concomitant with the shift from carbohydrate-based metabolism to utilization of fatty acids, which is known to follow food deprivation. Because fatty acids are more reduced than carbohydrates, their metabolism to CO 2 converts more NAD to NADH. Indeed, we show that the NAD/NADH ratio falls in pancreas of starved mice compared with fed controls, although we cannot be certain this change occurs specifically in islets, which represent a small percentage of pancreatic mass. A decrease in the NAD/NADH ratio has also been shown to down-regulate Sir2 activity in other physiological contexts [15,21,50].

UCPs and Longevity-Cellular Basis
The question arises whether the up-regulation of UCPs we find in b cells occurs in other tissues during food limitation or long-term CR. In a horizontal study of mice, a good correlation was found between the life span and the level of uncoupling of respiration [51]. It has also been shown that overexpression of a UCP increased the life span in Drosophila [52]. UCPs may confer benefit by damping production of reactive oxygen species during respiration. In one example, mitochondria from UCP3 KO mice generated higher levels of superoxide [53,54], suggesting that some degree of uncoupling can be beneficial by controlling reactive oxygen species production [55].
There are several reports that levels of UCPs may increase as a consequence of CR [56,57], but these findings are still not conclusive [58]. It will be interesting to see whether Sirt1 regulates UCPs in tissues other than b cells, and whether there is a broader change in the expression of UCPs during acute food limitation or CR.

Summary and Perspective
The induction of insulin secretion by glucose in pancreatic b cells has been a paradigm in mammalian cellular physiology. Here we describe a new level of regulation in which Sirt1 represses UCP2 to modulate the amplitude of insulin induction by glucose in b cells. We suggest that this mechanism serves to regulate chronic levels of insulin in accord with levels of food intake. This mechanism may reinforce the low levels of insulin production during food limitation and coordinate insulin release from b cells to the insulin sensitivity of metabolic tissues set by the diet. Regulation of UCP2 by Sirt1 may also be an important axis that is dysregulated by excess fat to contribute to obesityinduced diabetes.
Animal experimentation. The Sirt1 KO animals and their controls were obtained from M. McBurney [39]. The mice were housed under controlled conditions: temperature (25 6 1 8C) and light cycle ( Immunofluorescence. Pancreas was isolated from wild-type or Sirt1 KO mice and fixed in 4% paraformaldehyde. Consecutive 8-lm sections were immunostained with anti-Sirt1 antibody at 1:100 dilution (Upstate-Chemicon, Charlottesville, Virginia, United States). DAPI was used to stain the nuclei. Hematoxylin and eosin was used to visualize islets.
In vivo insulin measurement. Mice were subjected to O/N fast followed by intraperitoneal glucose injection (1 g/kg body weight). Blood samples were collected from the tail vein the night before the experiment (ad libitum), right before the glucose injection (time 0), and after the different time points indicated in Figure 2E. Insulin levels were measured using the Ultrasensitive Mouse Insulin EIA (Alpco Diagnostics) according to manufacturer's specifications.
Glucose tolerance test. Mice were fasted O/N (16 h) and injected intraperitoneally with a saline glucose solution at 1g/kg body weight. Plasma glucose levels were measured from tail blood before and 2, 5, 10, 25, 40, 80, and 120 min after the glucose injection. The test was repeated five times with results comparable to those in Figure 2D.
Western blot analysis. Cells were lysed in RIPA buffer (1% Chromatin immunoprecipitation. Assays were performed on cells as previously described according to the Farnham protocol (http:// mcardle.oncology.wisc.edu). Immunoprecipitations were performed with 1lg of Sirt1 antibody. PCR primers were designed to span a r e g u l a t o r y r e g i o n i n t h e U C P 2 p r o m o t e r ( 5 9-CGTCTGTTCAAAGCGTCTCA; 59-CCAGCTGGAGTCTTCTCCTT) (set 5) or a control region in the UCP2 promoter (59-AGGTTGTTTCTGGGCCATGTGCTCTAA; 59-TAGACCCTGGC CACCCTGAGCGCGAAAT) (set 4).
Islets isolation and morphometric analysis. Islets were isolated using a collagenase technique as described previously [59]. The pancreas was dissected and islets separated by incubation with collagenase at 37 8C. Islets were then hand-picked using a microscope. Islets were used for Western blot analyses or for glucose-induced insulin secretion assays. To measure insulin secretion, islets were incubated in KRB with 4 mM glucose for 1 h and then incubated with KRB containing different glucose concentrations (4-20 mM) for 1 h at 37 8C. The experiment was carried out twice with four mice per treatment, and the repeat experiment gave data comparable to Figure 2E. Four to five pancreases each from WT, Sir2þ/À, and Sir2À/À mice were fixed in 2% chilled paraformaldehyde, and paraffin embedded, and three sections (5 lm) separated by at least 100 lm were dewaxed using xylene, rehydrated through serial dilutions of ethyl-alcohol, and subjected to antigen retrieval using 10 mM citrate (pH 6.1), followed by DAKO high pH antigen retrieval solution (Dako, Glostrup, Denmark). The sections were washed and stained with the respective antibodies in staining buffer with 100 mM NaCl, 3% BSA, 0.5% Triton-X-100, and 50 mM Na-PO 4 (pH 7.4). Primary antibodies included sheep anti-insulin (The Binding Site, Birmingham, United Kingdom), mouse anti-glucagon (Sigma, St. Louis, Missouri, United States), and rabbit anti-somatostatin (Dako). The entire pancreatic section of each stained slide was arbitrarily divided into nonoverlapping contiguous portions (using a microscope inset measuring guide) and imaged at 103 magnification. The islet area (in square micrometers) and the total area of each section were determined by tracing the outline to assess the area of the islets, using image analyzer software (Image-Pro 4.1 Plus). The percentage of islet cell area in the pancreas was expressed as a percentage of the total pancreatic area. For 43 images, islets were stained using a cocktail of antibodies against insulin, glucagon, and somatostatin and HRPlinked secondary antibodies with DAB as a chromogen.
ATP/ADP measurements. 1 3 10 6 cells were washed with 13 PBS. Ice-cold 6% (v/v) HClO 4 was added, and immediately the cells were scraped from the plate. The solution was neutralized with 1 M KHCO 3 , centrifuged briefly, and the supernatant was passed through a 0.2-m filter (Nanosep, Lund, Sweden), and subjected to reversed phase chromatography using a Targa C18 250 3 4.6 mm 5-mm column as described in [61]. Nucleotides were detected at 260 nm with a Waters 486 tunable detector. Peak heights were measured. Nucleotide identities were confirmed by co-migration with known standards.
RNA analyses. Northern blot-Total RNA from INS-1 cells was extracted using Triazol reagent (Invitrogen). For Northern blot analysis, 10 lg of RNA samples was separated on a formaldehyde gel, transferred to nylon membrane, and then hybridized with genespecific probes. Normalization of mRNA was done by using actin as probe (kind gift from K. Olson). For whole-pancreas RNA extraction, fresh pancreas was stored in the RNA later solution (Ambion, Austin, Texas, United States) before extracting total RNA with the Qiagen system (Qiagen, Valencia, California, United States). The UCP2 probe was a kind gift from D. Ricquier, CNRS, Paris, France. RT-PCR/cDNA was synthesized from 1 lg of total RNA from the pancreas. The reverse transcriptase reaction was performed at 25 8C for 10  CAT assay. Transfections for CAT assays were done as described [60]. 293T cells were transiently transfected with pSU2N2-mUCP2-CAT (which contains 7 kb of the mouse UCP2 promoter; gift from D. Ricquier), pCMV-betaGal (to correct for transfection efficiency), and either pCMV-mSirt1 and pSPORT6-mPPAR-gamma2 (gift from B. Spiegelman, Harvard Medical School, Massachusetts, United States of America) or their respective empty vectors. 24 h later, total CAT enzyme levels were measured by ELISA (Roche, Basel, Switzerland).
NAD and NADH determination. NAD and NADH nucleotides were measured as described [15]. About 10 mg of frozen pancreas tissues was homogenized in 300 ll of acid extraction buffer to obtain NAD concentration, or alkali buffer to obtain NADH concentration. 240 ll of supernatant was neutralized with 120 ll of buffer. The concentration of nucleotides was measured fluorimetrically after an enzymatic cycling reaction using 2 ll of sample. All values were detected within the linear range and are expressed as nmol per gram of tissue.
Quantitation of NADH autofluorescence and glucose uptake. 1.5 3 10 6 INS-1 cells were seeded in six-well plates. Fresh medium containing 4.0 mM glucose was added the day before the experiment. The medium was removed on the day of the experiment, and the cells washed 33 with warm KRB buffer. The cells were then incubated with 1 ml of KRB at 37 8C for 1 h. At the end of the incubation, the media was removed and the cells were incubated for 10 min with KRB containing different glucose concentrations: 4-16.7 mM (for NADH determination), or 200 lM 2-NBDG (Molecular Probes) for glucose uptake determination. At the end of the incubation, cells were washed 33 with cold 13 PBS. We then used a four-laser LSR II digital flow cytometer (Becton-Dickinson, Mountain View, California, United States) with Diva software. The 488-mm laser excited the fluorescent glucose, and emission was detected with a 530/30BP filter and expressed as arbitrary units. NADH was measured according to Thorell [46] with a 355-mm laser and detected with a 440/40BP filter. Cell viability was assessed using PI. Data were analyzed with FlowJo (Tree Star, Ashland, Oregon, United States).