A Role for ATP-Citrate Lyase, Malic Enzyme, and Pyruvate/Citrate Cycling in Glucose-induced Insulin Secretion*

In pancreatic β-cells, metabolic coupling factors generated during glucose metabolism and pyruvate cycling through anaplerosis/cataplerosis processes contribute to the regulation of insulin secretion. Pyruvate/citrate cycling across the mitochondrial membrane leads to the production of malonyl-CoA and NADPH, two candidate coupling factors. To examine the implication of pyruvate/citrate cycling in glucose-induced insulin secretion (GIIS), different steps of the cycle were inhibited in INS 832/13 cells by pharmacological inhibitors and/or RNA interference (RNAi) technology: mitochondrial citrate export, ATP-citrate lyase (ACL), and cytosolic malic enzyme (ME1). The inhibitors of the di- and tri-carboxylate carriers, n-butylmalonate and 1,2,3-benzenetricarboxylate, respectively, reduced GIIS, indicating the importance of transmitochondrial transport of tri- and dicarboxylates in the action of glucose. To directly test the role of ACL and ME1 in GIIS, small hairpin RNA (shRNA) were used to selectively decrease ACL or ME1 expression in transfected INS 832/13 cells. shRNA-ACL reduced ACL protein levels by 67%, and this was accompanied by a reduction in GIIS. The amplification/KATP-independent pathway of GIIS was affected by RNAi knockdown of ACL. The ACL inhibitor radicicol also curtailed GIIS. shRNA-ME1 reduced ME1 activity by 62% and decreased GIIS. RNAi suppression of either ACL or ME1 did not affect glucose oxidation. However, because ACL is required for malonyl-CoA formation, inhibition of ACL expression by shRNA-ACL decreased glucose incorporation into palmitate and increased fatty acid oxidation in INS 832/13 cells. Taken together, the results underscore the importance of pyruvate/citrate cycling in pancreatic β-cell metabolic signaling and the regulation of GIIS.

Pancreatic ␤-cells secrete insulin in response to an elevated blood glucose level. Although the molecular aspects of glucosestimulated insulin secretion are not completely understood, it is clear that an increase in the cytosolic ATP to ADP ratio plays a critical role in this process via the closure of ATP-dependent potassium (K ATP ) 3 channels, leading to the opening of voltagegated Ca 2ϩ channels (1)(2)(3). The resulting Ca 2ϩ influx stimulates insulin vesicules exocytosis (4). In recent years, in addition to K ATP /Ca 2ϩ signaling, metabolic coupling factors (MCFs) derived from anaplerosis (e.g. malonyl-CoA (11,12), succinate (5), and glutamate (6)), NADPH (7,8), H 2 O 2 (9), lipolysis of endogeneous lipid stores (10,11), and lipid signaling molecules, in particular free fatty acids themselves, fatty acyl-CoAs and diacylglycerol (12)(13)(14)(15)(16)(17)(18), have been proposed to be critical players in fuel-stimulated insulin secretion by ␤-cells. However, the precise mechanisms underlying the amplification (formerly called K ATP -independent) pathways of fuel-induced insulin secretion and the targets of the various candidate MCFs are poorly understood.
Anaplerosis, the replenishment of the tricarboxylic acid cycle with intermediates, is essential for its optimal operation in different cell types and is involved in the regulation of glucose and lipid metabolisms (review in Ref. 19). The anaplerotic enzyme pyruvate carboxylase (PC) is highly expressed in pancreatic ␤-cells (7,20) and in rat clonal INS-1 cells (21). Nearly 50% of the glucose carbon entering the mitochondria is utilized by PC and converted to oxalacetate in rat islet tissue (22,23). Considering the absence of gluconeogenesis (24) and the relatively low rate of lipid synthesis (21) in the ␤-cells, the high PC activity suggests an alternate purpose for anaplerosis in the ␤-cell. We and others have investigated the importance of anaplerosis in glucose-induced insulin secretion (GIIS) (20,(25)(26)(27) and have shown that following accelerated anaplerosis in glucose-stimulated ␤-cells, different tricarboxylic acid cycle intermediates, such as citrate and malate (20,27), are exported from the mitochondria to the cytosol (cataplerosis) where they could play a signaling role in GIIS.
Three different mitochondrial metabolite shuttles/cycling processes linking anaplerosis/cataplerosis with GIIS have been described (reviewed in Ref. 28). These are: 1) the pyruvate/ malate shuttle (7), 2) pyruvate/citrate cycling (27), and 3) the pyruvate/isocitrate/␣-ketoglutarate shuttle (29). A common * This work was supported in part by a grant from the Canadian Institute of Health Research (to M. P.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. 1  feature of these shuttles is the formation of NADPH, a candidate MCF (7,8). However, among these shuttles, only pyruvate/ citrate cycling leads, in addition, to the production of malonyl-CoA, another candidate MCF (14) and is associated with the reoxidation of cytosolic NAD, possibly favoring fast glycolytic flux (23). As part of this cycling process, citrate derived from anaplerotic input into the tricarboxylic acid cycle is exported from mitochondria to the cytosol, where it is cleaved by ATPcitrate lyase (ACL) to acetyl-CoA and oxaloacetate.  (20), ME1 activity is thought to be the major source of NADPH in the ␤-cell, at least in human and rat islets, but perhaps not in mouse islets that might contain a very low level of the enzyme (30). Finally, recent studies have shown that inhibition of citrate export lowers GIIS (31) and that a decrease in ME1 expression alters glucose-and amino acid-induced insulin secretion (32).
In the present study, we continued our investigation of the role of pyruvate/citrate cycling in the regulation of GIIS using two complementary approaches in INS 832/13 cells. We sought to directly test the hypothesis that this particular pyruvate cycling process is important for GIIS. First, pharmacological inhibitors were used to block citrate export out of the mitochondria and to inhibit ACL. Second, shRNAs were designed to specifically decrease the expression levels of ACL and ME1. The results show that a reduction in mitochondrial citrate export, ACL inhibition, or a reduction in ACL and ME1 expression all lead to a decrease in GIIS. Altogether the data provide direct support for a role of pyruvate/citrate cycling in GIIS.

EXPERIMENTAL PROCEDURES
Cell Culture-INS 832/13 cells (33) (passages 51-64) were grown in monolayer cultures in RPMI 1640 complete medium at 11.1 mmol/liter glucose supplemented with 10% (w/v) fetal bovine serum, 10 mmol/liter HEPES, 2 mmol/liter L-glutamine, 1 mmol/liter sodium pyruvate and 50 mol/liter ␤-mercaptoethanol at 37°C in a humidified atmosphere (5% CO 2 , 95% air). For transfection experiments, the cells were seeded in 75-cm 2 flasks at 4 ϫ 10 6 cells 2 days prior to transfection and were at 60 -70% confluency at the time of the transfection. For pharmaceutical inhibitor experiments, the cells were seeded in a 12-well plate at a density of 2 ϫ 10 5 cells/well. Short Hairpin RNA-mediated Gene Suppression-shRNAs directed against rat cytosolic NADP ϩ -dependent malic enzyme (ME1; GenBank TM accession number NM_012600) or rat ACL (GenBank TM accession number NM_016987) were designed according to Ambion (Austin, TX) siRNA design guidelines. Briefly, shRNA sequence included, from 5Ј to 3Ј, BamHI restriction site (in bold type), the 19-nucleotide sense genetargeting sequence (underlined), loop sequence (in italics), the 19-nucleotide antisense gene-targeting sequence (underlined), TTT TTT ending transcription sequence, and HindIII restriction site (in bold type). For each targeted-gene, four different shRNAs were designed and tested for their potency to decrease the targeted gene expression. The most efficient one was chosen for subsequent experiments and its sequence is listed here. The sense sequence for shRNA-ME1 (beginning at nucleotide 387 of ME1 sequence) is 5Ј-GAT CCG GGC ATA TTG CTT CAG TTC TTC AAG AGA GAA CTG AAG CAA TAT GCC CTT TTT TGG AAA-3Ј, and that for shRNA-ACL (begining at nucleotide 560 of ACL sequence) is 5Ј-GAT CCG ACA TTA AGA GAC ACC TGT TTC AAG AGA ACA GGT GTC TCT TAA TGT CTT TTT TGG AAA-3Ј. The sense sequences with no known target for appropriate scrambled control were 5Ј-GAT CCC AGT GCT GGT ACT TGT ACT TTC AAG AGA AGT ACA AGT ACC AGC ACT GTT TTT TGG AAA-3Ј and 5Ј-GAT CCC TGA GAA TCT AAC GTA AGC TTC AAG AGA GCT TAC GTT AGA TTC TCA GTT TTT TGG AAA-3Ј for Scr-ME1 and Scr-ACL, respectively. All of the oligonucleotides were synthesized and purified by polyacrylamide gel electrophoresis by IDT (Coralville, IA). After annealing, the doublestranded oligonucleotides were ligated by self-formed restriction sites for BamHI and HindIII in pSilencer 2.0 vector (Ambion). shRNAs expression was under U6 promotor control. The pSilencer 2.0 empty vector (referred to as "mock" in the text) was also used as a control in most of the experiments.
Real Time Quantitative PCR Analysis-The cells were extracted using the RNeasy mini kit (Qiagen) with RNase-free DNase (Qiagen) from 12-well plates. RNA was reverse transcripted to cDNA using Moloney murine leukemia virus reverse transcriptase (Invitrogen) and hexamers as described previously (34). Forward and reverse sequence primers (IDT) were designed specifically for ME1 or ACL and normalized to cyclophilin. Primer sequences were: ME1 forward, 5Ј-CCG CAT CTC AAC AAG GAC TTG-3Ј, and reverse, 5Ј-CCA TCA GTC ACC ACA ATA GCC-3Ј; ACL forward, 5Ј-TGC TTA GCC AGA GCT TGG TAG-3Ј, and reverse, 5Ј-TGA ATA GGC CGG AGA TGA AGC-3Ј; cyclophilin forward, 5Ј-CTT GCT GCA GAC ATG GTC AAC-3Ј, and reverse, 5Ј-GCC ATT ATG GCG TGT GAA GTC-3Ј. Real time quantitative PCR were performed on Rotor-Gene RG-3000 (Corbett Research, Mortlake, Australia) using LC Faststart DNA Master plus SYBR Green reagent (Roche Applied Science). The results were analyzed using the Rotor-Gene software version 6.0.19 provided by Corbett Research.
Cytosolic NADP ϩ -dependent Malic Enzyme 1 Activity-Malic enzyme activity was assayed in fresh cytosolic cell extracts. Briefly, 96 h post-transfection, shRNA-ME1, ScrME1, or mock transfected cells were placed on ice, washed with cold PBS, extracted by scraping in ice-cold PBS/EGTA buffer containing protease inhibitors (1 g/ml of leupeptin and pepstatin, 1 mmol/liter phenylmethylsulfonyl fluoride, and 1.5 g/ml of aprotinin and referred to as protease inhibitor mix in the text). Cell suspension was centrifuged at 1000 ϫ g for 2 min at 4°C. The cell pellet was resuspended in 500 l in a ice-cold homogenization buffer containing 20 mmol/liter HEPES, pH 7.4, 0.25 mol/liter sucrose, 1 mmol/liter EGTA, and protease inhibitor mix and homogenized using a 1-ml Potter-Elvehjem homogenizer. The homogenates were centrifuged at 1000 ϫ g for 2 min at 4°C to discard nucleus and then at 12,000 ϫ g for 30 min at 4°C to discard mitochondria. The post-mitochondrial supernatant content was concentrated using a Microcon YM-30 (Millipore, Billerica, MA). Final volume of the concentrate was adjusted to 100 l with ice-cold homogenization buffer (see above), and 40 l was used for malic enzyme assay. The assay system in 96-well plates contained 70 mmol/liter of triethanolamine-HCl buffer, pH 7.4, 5 mmol/liter MnCl 2 , 0.3 mmol/liter NADP ϩ , 3 mmol/liter of L-malate. The reactions were started adding cytosolic extracts immediately followed by shaking, and absorbance at 340 nm was read every 30 s for up to 10 min using Fluostar Optima (BMG Labtechnologies, Offenburg, Germany). For each set of experiments, background control was run without L-malate as substrate. Enzyme activity was determined by subtracting the activity of the background control to each sample. The resulting slopes of absorbance versus time were normalized to protein content using a BCA protein kit (Pierce).
Immunoblot Analysis-shRNA-ACL, ScrACL, or mock transfected cells were collected at 24, 48, 72, and 96 h posttransfection. The cells were rinsed with cold PBS, trypsinized, and centrifuged at 300 ϫ g for 5 min at 4°C. The cell pellets were washed once with cold PBS, resuspended in 20 mmol/liter Tris-HCl lysis buffer containing 150 mmol/liter NaCl, 1 mmol/ liter EDTA, 1 mmol/liter EGTA, 1% (v/v) Triton X-100, 0.1% SDS, supplemented with protease inhibitor mix, incubated on ice for 45 min, and centrifuged 10 000 ϫ g for 10 min at 4°C. The supernatant containing total cell extract was collected, and the protein content was determined using a BCA protein assay kit (Pierce). Aliquots of 20 g of total cell extract were resolved on a 8% SDS-PAGE and transferred to nitrocellulose membrane (Whatman, Hahnestrabe, Dassel, Germany). The membranes were incubated with anti-ACL antibody (Cell Signaling Technology, Danvers, MA) for 1 h at room temperature, followed by secondary antibody treatment with horseradish peroxidase-conjugated goat anti-rabbit IgG (Bio-Rad). As a load-ing control, the membranes were incubated with anti-tubulin antibody (Abcam Inc., Cambridge, MA). Semi-quantitative analysis of immunoblot density was performed using the Genesnap software from the G-Box (PerkinElmer Life Sciences).
Insulin Secretion and Insulin Content Measurement-shRNA-transfected cells were preincubated during 2 h in RPMI complete medium at 1 mmol/liter glucose and then were equilibrated for 1 h in Krebs-Ringer bicarbonate buffer containing 10 mmol/liter HEPES (KRBH, pH 7.4), 0.5% defatted BSA (Sigma-Aldrich), and 1 mmol/liter glucose. The cells were then incubated for 45 min in KRBH containing 0.5% BSA and 1, 5 or 10 mmol/liter glucose, in the presence or absence of 0.3 mmol/ liter palmitate, or 35 mmol/liter KCl, and/or 20 mol/liter diazoxide. Stock solution of diazoxide was prepared in Me 2 SO, and the final Me 2 SO concentration in KRBH was 0.5%. In experiments using inhibitors, INS 832/13 cells were incubated as described above in the presence or absence of 0.5 mmol/liter n-butylmalonate (BM) (Sigma) or 1,2,3-benzenetricarboxylate (BTC) (Sigma) or 50 M radicicol (Sigma). Stock solution of radicicol was prepared in Me 2 SO, and the final Me 2 SO concentration in KRBH was 0.2%. At the end of incubation, the media were preserved for insulin determination. Total insulin content of the cells was measured after acid-ethanol (0.2 mmol/liter HCl in 75% ethanol) extraction. Insulin levels were measured by radioimmunoassay using a human insulin standard (Linco Research, St. Charles, MO). The insulin levels were normalized to total insulin content and to protein content.
Glucose and Fatty Acid Oxidation and Glucose Incorporation into Fatty Acids-Glucose oxidation to CO 2 , fatty acid (FA) oxidation to CO 2 and acid-soluble products (essentially ketone bodies (35)), and glucose incorporation into FA were measured in transfected cells preincubated as described for insulin secretion assay. For glucose oxidation experiments, the cells were incubated for 45 min in KRBH containing 0.5% BSA, 1 mmol/ liter glucose, and 0.1 Ci/ml of [U-14 C]glucose (Amersham Biosciences) or 10 mmol/liter glucose and 0.2 Ci/ml of [U-14 C]glucose. At the beginning of the incubation, the 25-cm 2 flasks were sealed with a stopper supporting a 3-cm length of PVC tubing (internal diameter, 4.7 mm) containing a piece of Whatman GF/B paper (half of a 25-mm-diameter circle) soaked in 5% KOH. At the end of incubation period, 0.2 ml of perchloric acid (40% v/v) was injected into each flask via a needle through the cap to acidify the medium and to liberate the CO 2 . After overnight isotopic equilibration at room temperature, the papers were removed, and the trapped 14 CO 2 was measured by liquid scintillation counting after overnight equilibration in scintillation fluid. For fatty acid oxidation experiments, the cells were incubated for 45 min in KRBH containing 0.5% BSA, 1, 5, or 10 mmol/liter glucose in the presence of 1 mmol/liter of carnitine plus 0.2 mmol/liter of palmitate, and 0.1 Ci/ml of [1-14 C]palmitate (PerkinElmer Life Sciences). Oxidation to CO 2 and acid-soluble products were performed as previously described (17). For glucose incorporation into FA experiments, the cells were incubated for 2 h in KRBH containing 0.5% BSA, 1, 5, or 10 mmol/liter glucose, and 3 Ci/ml [U-14 C]glucose (Amersham Biosciences) in the presence of 1 mmol/liter oleate. At the end of the incubation period, the media were removed, the cells were washed once with PBS, and 1 ml of PBS with 0.53 mmol/liter EDTA were added to the plated cells. The cells were collected with gentle pipetting, centrifuged at 500 ϫ g for 10 min at 4°C, dissolved in 3 ml of CHCL 3 :methanol:HCl (200: 100:1 v/v), and placed at Ϫ20°C until the extraction. To perform the lipid extraction, 750 l of H 2 O was added, and after vigorous vortexing, the phases were separated by centrifugation at 1 000 ϫ g for 10 min at 4°C. The organic lipid layer was extracted and evaporated under a steam of N 2 to dryness. To separate [ 14 C]FA and [ 14 C]glycerol formed from [ 14 C]glucose, esterification products were hydrolyzed for 1 h at 37°C in 100 l of 0.1 mmol/liter of Tris buffer, pH 7.0, containing 0.5% of Triton X-100, 1% of acetonitrile, Ͼ30 units of both lipase and phospholipase C (Sigma) enzymes. At the end of incubation, 3 ml of chloroform:methanol:HCl (200:100:1 v/v/v) was added to the hydrolysis buffer. The lipids were re-extracted as described above and redissolved in 50 l of chloroform and spotted onto a silica gel 60 TLC plate (Whatman, Florham Park, NJ). Extracted lipids were separated by thin layer chromatography using silica gel 60 plates in a solvent system containing petroleum ether:ethyl ether:acetic acid (70:30:1). Radioactive signal present on the TLC plate was transferred to a storage phosphor screen (Amersham Biosciences) and visualized after scanning of the screen on the Typhoon (Amersham Biosciences). Semiquantitative analysis of [ 14 C]FA (including FA from esterification products and very low level of NEFA) was performed, using [ 14 C]palmitate standard as a reference, with ImageQuant TL software (Amersham Biosciences) and quantification of [ 14 C]glucose incorporated into FA was performed using the [ 14 C]glucose standard curve.
Statistical Analysis-The data are expressed as the means Ϯ S.E. Statistical significance was calculated with two-tailed unpaired Student's t test. A p value of Ͻ0.05 was considered significant (*), whereas a p value of Ͻ0.01 was considered very significant (**). The statistical analyses were performed using the InStat program (GraphPad Software, San Diego, CA).

Inhibition of Mitochondrial Citrate Export Alters GIIS-We
have previously reported that an elevation in glucose concentration causes a rise in citrate in both mitochondrial and cytosolic compartments in the ␤-cell (27), suggesting that during glucose stimulation, citrate formed by the tricarboxylic acid cycle is exported from the mitochondria into the cytosol. Both the di-and tri-carboxylate carriers are essential for effective exportofcitrate (36,37).Toinvestigatetheimplicationofpyruvate/ citrate cycling in GIIS, we first decreased mitochondrial citrate export using BM and BTC, inhibitors of the di-and the tricarboxylate carriers, respectively. As shown in Fig. 1, incubation of INS 832/13 cells with 0.5 mM BM or BTA decreased insulin secretion at high glucose (10 mM). BTA also reduced insulin secretion at intermediate glucose (5 mM), whereas BM had no effect. KCl-stimulated insulin secretion was not affected by BM or BTA, and the cellular insulin content was unaltered by these treatments (data not shown). These results suggest that export of citrate from mitochondria is needed for GIIS.

Inhibition of ATP-Citrate Lyase by Radicicol Decreases GIIS-
The next step in the pyruvate/citrate cycling process involves cytosolic citrate cleavage by ACL to acetyl-CoA and oxaloacetate. To further examine the role of pyruvate/citrate cycling in the regulation of GIIS, we decided to study the importance of ACL by using specific inhibitors of this enzyme. Hydroxycitrate (38 -40) and radicicol (38) have been used as ACL inhibitors, and their effects on GIIS have been investigated. However, contradictory results have been published using hydroxycitrate in rat islets (38 -40). Because hydroxycitrate decreases the pH of KRBH buffer and is known to actively bind Ca 2ϩ , we did not test this inhibitor. Radicicol, a noncompetitive inhibitor of ACL (41), was used instead to determine the effect of ACL inhibition on GIIS in INS 832/13 cells. In accordance with a previous study carried out in purified normal rat ␤-cells (38), 50 M radicicol decreased insulin secretion by ϳ50% at 5 and 10 mM glucose but had no effect on KCl-stimulated insulin release compared with control (Fig. 2). In all of the experiments, basal insulin secretion at 1 mM glucose and the total cellular insulin content  were not affected by radicicol (data not shown). Radicicol also appeared to decrease GIIS in the presence of 0.3 mM palmitate, although the effect was less prominent and did not reach statistical significance (Fig. 2).
shRNA Against ACL Decreases Its Expression and Inhibits GIIS-The RNA interference technology was used as a complementary, nonpharmacological, approach. Four shRNA constructs specifically directed against ACL (shRNA-ACLs 1-4) were tested for their efficiency to decrease ACL expression, as compared with a scrambled control shRNA. All four shRNA-ACL constructs reduced ACL mRNA level in INS 832/13 cells from 35 to 75%, 48 h post-transfection (data not shown). The most efficient, shRNA-ACL 3, was used in subsequent experiments and is referred to shRNA-ACL. We next determined the time-dependent efficiency of shRNA-ACL at 14, 24, 36, 48, 72, and 96 h post-transfection. shRNA-ACL decreased ACL mRNA level as early as 14 h (50 Ϯ 0.5% reduction versus t ϭ 0) with a maximal effect between 36 and 48 h post-transfection (74 Ϯ 2 and 75 Ϯ 1.5% versus t ϭ 0, respectively). At the protein level, ACL expression was reduced by 30% at 24 h and by 75% at 96 h, as compared with ScrACL. Because suppression of ACL expression for a long period of time could affect cell growth (42,43), we decided to perform the experiments at 48 h post-transfection.
Insulin secretion was assayed in INS 832/13 cells transfected with shRNA-ACL, the scrambled control (ScrACL), or the empty vector (mock). At 10 mM glucose, but not 5 mM, the insulin secretion in ScrACL-treated cells was slightly but significantly lower (p Ͻ 0.05), as compared with empty vector-transfected cells (mock). Perhaps the insert with shRNA-ScrACL very slightly perturbed the insulin release machinery, although there was no difference at 10 mM glucose plus palmitate or 5 mM glucose plus or minus palmitate. In addition, we verified that the shRNA-ScrACL did not affect basal and KCl-induced insulin secretion as well as glucose oxidation, fatty acid oxidation, and ACL expression. Knockdown of ACL protein expression by 67% with shRNA-ACL (see Fig. 3A) significantly inhibited GIIS at both intermediate (5 mM, 43% versus ScrACL) and elevated (10 mM, 25% versus ScrACL) glucose concentrations (see Fig.  3B) without affecting the secretory effect of KCl. Basal insulin release at 1 mM glucose (Fig. 3B) and the cellular insulin content (6.0 Ϯ 0.3 and 6.0 Ϯ 0.4 ng of insulin/g of protein for ScrACL and shRNA-ACL, respectively; n ϭ 4) were not affected by the reduction in ACL expression. In the presence of 0.3 mM palmitate, ACL knockdown in INS 832/13 cells led to a 27% decrease in GIIS at 5 mM glucose, whereas at 10 mM glucose the reduction in GIIS was not statistically significant (Fig. 3B). The fact that reduced insulin secretion by either radicicol or ACL knockdown was less prominent when the experiments were performed in the presence of exogenous fatty acids fits with the concept that pyruvate/citrate cycling is linked to the production of lipid signaling molecules, e.g. malonyl-CoA and fatty acyl-CoA.
To further understand the role of ACL in the regulation of GIIS, we investigated the effect of ACL knockdown on the amplification/K ATP -independent pathway. The combination of diazoxide (an activator of the K ATP channel) with a stimulatory concentration of elevated KCl allows the assessment of GIIS at high cytosolic Ca 2ϩ by a pathway independent of K ATP channel closure (44). As shown in Fig. 3C, insulin secretion at 5 and 10 mM glucose in presence of diazoxide and KCl was decreased by 20 and 37%, respectively, as compared with ScrACL control, indicating that ACL activity is implicated in metabolic signaling by glucose and is needed for the optimal operation of the K ATPindependent pathway(s) of GIIS.
ACL Down-regulation Affects Fatty Acid Metabolism but Not Glucose Oxidation-Acetyl-CoA produced from citrate cleavage by ACL is converted to malonyl-CoA, a candidate MCF. Prentki et al. (14) have suggested that malonyl-CoA by inhibiting carnitine palmitoyl transferase 1 (CPT-1), decreases FA oxidation and diverts fatty acyl-CoA into esterification products, leading to a potentiation effect of FA on GIIS. To test whether the decrease in GIIS observed using shRNA-ACL could be linked to an alteration in FA metabolism, FA oxidation was measured using [ 14 C]palmitate. In both empty vector (mock) and ScrACL transfected INS 832/13 cells, increasing glucose concentration in the incubation medium led to a dose-dependent decrease in FA oxidation (Fig. 4A). This decrease in FA oxidation in response to rising glucose was also observed in shRNA-ACL transfected cells, but at a lesser extent. As shown in Fig. 4A, FA oxidation in shRNA-ACL transfected cells was increased at intermediate (5 mM) and elevated (10 mM) glucose concentration compared with the ScrACL or mock controls. We also investigated whether an inhibition of ACL expression affects FA synthesis and elongation, a process that requires malonyl-CoA and NADPH (two candidate MCF), by measuring [ 14 C]glucose incorporation into FA. Knockdown of ACL expression by shRNA-ACL reduced glucose incorporation into FA at both 5 and 10 mM glucose (Fig. 4B). Glucose oxidation measured in cells transfected with shRNA-ACL did not differ from control conditions (mock and ScrACL) at either basal (1 mM) or elevated (10 mM) glucose (Fig. 4C). Collectively these results suggest that knockdown of ACL expression by shRNA-ACL did not affect glucose oxidation but did alter FA metabolism, possibly via a reduction in malonyl-CoA and/or NADPH levels.
shRNA-ME1 Decreases ME1 Expression and Inhibits GIIS-Finally to ascertain whether cytosolic ME (ME1), which catalyzes the last enzymatic step in the pyruvate/citrate cycling process, contributes to the regulation of GIIS, we knocked down ME1 by shRNA in INS 832/13 cells. We designed and tested four different shRNA specifically directed against ME1 (shRNA -ME1s 1-4). Compared with a scrambled control (ScrME), shRNAs 1-4 decreased ME1 expression at the mRNA level by 40 -60% 24 h post-transfection (data not shown). The most efficient, shRNA-ME1 3 was chosen for subsequent experiments and is mentioned as shRNA-ME1. Determination of the time-dependent efficiency of shRNA-ME1 showed that shRNA-ME1 decreased ME1 expression at the mRNA level after 24 h (42% versus t ϭ 0) with a maximal effect between 36 and 48 h post-transfection (81%). However, ME1 activity was decreased only by less than 50% at 48 and 72 h post-transfection. We therefore performed the experiments 96 h post-transfection, when ME1 activity was reduced by more than 60% (Fig.  5A). No sign of toxicity (assayed by trypan blue; data not shown) was observed in INS 832/13 cells transfected with shRNA-ME1, ScrME, or empty vector (mock).
Knockdown of ME1 activity decreased insulin secretion at intermediate (5 mM) and elevated (10 mM) glucose by more than 30% compared with ScrME, without any effect on KClstimulated insulin release (Fig. 5B). Basal insulin release at 1 mM glucose (Fig. 5B) and the cellular insulin content were not affected by the reduction in ME1 expression (3.9 Ϯ 0.3 ng of insulin/g of protein and 4.0 Ϯ 0.4 ng of insulin/g of protein for ScrME and shRNA-ME1, respectively; n ϭ 3). Oxidative metabolism of glucose was not affected by shRNA-ME1 (Fig. 5C).

DISCUSSION
In a previous study, we showed that a pyruvate/citrate shuttle is operational in INS-1 ␤-cells and provided correlative biochemical evidence suggesting that anaplerosis/cataplerosis is implicated in fuel-induced insulin secretion (27). The current study aimed at directly testing these hypotheses using pharmacological and molecular approaches. As detailed below through the examination of each step in the cycle, the current results, together with our previous work (20,21,27), provide strong evidence supporting the view that pyruvate/citrate cycling is implicated in the amplification/K ATP -independent pathway of glucose signaling for insulin secretion.
The first step of this cycling process involves the carboxylation of pyruvate by PC (Fig. 6). We previously showed that inhibiting this enzyme with phenylacetate curtailed GIIS and that the dose dependence of the drug inhibitory effect on insulin secretion correlated with reductions in cellular citrate (23). In addition, glucose stimulation of the ␤-cell was associated  with a rise in both mitochondrial (corresponding to the result of the second reaction in the cycle catalyzed by citrate synthase) and cytosolic citrate (23). The implication in glucose signaling of the export of citrate from the mitochondria to the cytosol (cataplerotic output), the third step of pyruvate/citrate cycling, has been studied by Joseph et al. (31). These authors recently reported that either pharmacological inhibition of the tricarboxylate carrier (TIC) using BTC or TIC knockdown using specific siRNA resulted in a reduction of GIIS in INS 832/13 cells and rat islets. However, even though TIC conducts citrate export from the mitochondria (37), the dicarboxylate carrier (DIC) supplies malate for the exchange transport of citrate through TIC (36,45). In the current study, the inhibition of either TIC (with BTC) or DIC (with BM) resulted in a decrease in GIIS. These results suggest that both di-and tri-carboxylate carriers are important for glucose signaling. Once in the cytosol, citrate is cleaved by ACL to oxaloacetate and acetyl-CoA, which can be converted to malonyl-CoA. In the present study, radicicol, a pharmacological inhibitor of ACL, decreased GIIS in INS 832/13 cells, similar to earlier results in dispersed rat pancreatic ␤-cells (38), suggesting that inhibition of ACL activity is important for GIIS. The involvement of ACL in GIIS was further confirmed using the RNA interference approach. Knockdown of ACL expression by shRNA-ACL reduced GIIS and affected the K ATP -independent pathway of GIIS. Because ACL activity provides substrate for the formation of malonyl-CoA, an inhibitor of CPT-1 and FA oxidation, we verified that knockdown of ACL led to a rise in FA oxidation. Because glucose-derived cytosolic malonyl-CoA, and NADPH generated by the cycle are utilized for FA synthesis, we also verified the prediction that ACL knockdown does lead to a reduction in [ 14 C]glucose incorporation into FA. The penultimate step in pyruvate/citrate cycling involves the conversion of malate (derived from the reduction of oxaloacetate by malate dehydrogenase) into pyruvate by the cytosolic isoform of ME (ME1). Knockdown of ME1 decreased GIIS without affecting glucose oxidation. These results implicate ME1 in glucose-induced insulin secretion in INS 832/13 cells. Similar results were described previously by Pongratz et al. (32). Finally, the last step in the cycle involves the mitochondrial pyruvate transporter, and others showed that inhibiting this carrier with ␣-cyano-4-hydroxycinnamate curtailed GIIS in the rat pancreatic islets (46). Collectively, the above results provide very strong support for the concept that pyruvate/citrate cycling plays an important role in GIIS.  Three different pyruvate cycling processes (pyruvate/malate, pyruvate/citrate, and pyruvate/isocitrate/␣-ketoglutarate have been proposed to be active in the ␤-cell, possibly linking mitochondrial metabolism to insulin secretion. It should be underscored that none of these pyruvate cycling processes are exclusive and that they might be redundant or vary in activity depending on species or the nutritional state. A common feature of these metabolite shuttles is the production of NADPH. Ivarsson et al. (8) have shown that elevation in glucose concentration correlates with a rise in NADPH to NADP ϩ ratio and ␤-cell capacitance, a measurement for insulin granule exocytosis. This supported earlier suggestions of Hedeskov et al. (47) that the cytosolic NADPH to NADP ϩ ratio, rather than the NADH to NAD ϩ ratio, establishes a link between glucose metabolism and insulin secretion. In pancreatic ␤-cells, most of the cytosolic NADPH is produced by ME1 and/or cytosolic isocitrate dehydrogenase, but little is produced by glucose-6 phosphate dehydrogenase in the pentose phosphate pathway, because a relatively low amount of glucose is metabolized through this pathway (20). Whether NADPH acts directly or indirectly on insulin exocytosis remains to be determined. NADP(H) could act on insulin release by binding to the K v ␤-subunits of the voltage-dependent K ϩ channels (48) or perhaps by binding to the nucleotide inhibitory site of the K ATP channel (49). Different NADPH-using enzymes have been reported to regulate insulin secretion (for review see Ref. 28), including thioredoxin and glutaredoxin (8) and the neuronal isoform of nitric-oxide synthase (50,51), which are additional candidates for the possible role of NADPH in the regulation of insulin release.
Among the three pyruvate shuttles linking glucose metabolism to insulin secretion, only pyruvate/citrate cycling contributes to malonyl-CoA formation and the reoxidation of cytosolic NAD. These two features render this cycling process particularly attractive with respect to the two others because both features favor high glycolytic flux. Thus, the NAD regeneration allows flux through glyceraldehyde-3-phosphate dehydrogenase, whereas high malonyl-CoA, by inhibiting fatty acid oxidation, should allow full glucose usage and also prevents a possible inhibition of pyruvate dehydrogenase by fatty acid oxidation and the establishment of the Randle cycle.
The hypothesis presented here that pyruvate/citrate cycling is important for GIIS is attractive, not only because this cycling process is linked to the production of NADPH, NAD ϩ and malonyl-CoA in the cytoplasm, but also because enhanced pyruvate/citrate cycling might be related to mitochondrial ATP production (Fig. 6). Thus, during the operation of the pyruvate/ citrate shuttle, malate is co-transported by both the TIC and DIC carriers with a net influx of inorganic phosphate. The latter is a substrate for ATP synthase, and ATP is released in the cytoplasm through adenine nucleotide translocase. Thus, in response to an elevation in glucose concentration, activation of mitochondrial metabolism and of pyruvate cycling through the pyruvate/citrate shuttle should favor the net synthesis of ATP, malonyl-CoA, and NADPH, three key candidate MCFs for the regulation of insulin secretion, as well as cytosolic NAD ϩ , important for the positive regulation of the glycolytic flux. Moreover, Fransson et al. (52) have recently provided evidence that PC, and thus anaplerosis, is important for the rise of ATP to ADP ratio in response to glucose.
In conclusion, the results support the hypothesis that pyruvate/citrate cycling plays a critical role in the regulation of insulin secretion in response to glucose. Because pyruvate cycling, possibly via a rise in PC and ME1 expression, is increased in rodent models of ␤-cell compensation for insulin resistance (26,53) in which insulin secretion is enhanced, it will also be of interest to determine whether pyruvate/citrate cycling plays a role in ␤-cell compensation and failure in the etiology of type 2 diabetes.