Transamination Is Required for α-Ketoisocaproate but Not Leucine to Stimulate Insulin Secretion*

It remains unclear how α-ketoisocaproate (KIC) and leucine are metabolized to stimulate insulin secretion. Mitochondrial BCATm (branched-chain aminotransferase) catalyzes reversible transamination of leucine and α-ketoglutarate to KIC and glutamate, the first step of leucine catabolism. We investigated the biochemical mechanisms of KIC and leucine-stimulated insulin secretion (KICSIS and LSIS, respectively) using BCATm−/− mice. In static incubation, BCATm disruption abolished insulin secretion by KIC, d,l-α-keto-β-methylvalerate, and α-ketocaproate without altering stimulation by glucose, leucine, or α-ketoglutarate. Similarly, during pancreas perfusions in BCATm−/− mice, glucose and arginine stimulated insulin release, whereas KICSIS was largely abolished. During islet perifusions, KIC and 2 mm glutamine caused robust dose-dependent insulin secretion in BCATm+/+ not BCATm−/− islets, whereas LSIS was unaffected. Consistently, in contrast to BCATm+/+ islets, the increases of the ATP concentration and NADPH/NADP+ ratio in response to KIC were largely blunted in BCATm−/− islets. Compared with nontreated islets, the combination of KIC/glutamine (10/2 mm) did not influence α-ketoglutarate concentrations but caused 120 and 33% increases in malate in BCATm+/+ and BCATm−/− islets, respectively. Although leucine oxidation and KIC transamination were blocked in BCATm−/− islets, KIC oxidation was unaltered. These data indicate that KICSIS requires transamination of KIC and glutamate to leucine and α-ketoglutarate, respectively. LSIS does not require leucine catabolism and may be through leucine activation of glutamate dehydrogenase. Thus, KICSIS and LSIS occur by enhancing the metabolism of glutamine/glutamate to α-ketoglutarate, which, in turn, is metabolized to produce the intracellular signals such as ATP and NADPH for insulin secretion.

To maintain glucose homeostasis in response to a meal, insulin secretion is precisely stimulated by nutrients such as glu-cose, amino acids, and free fatty acids as well as incretin hormones such as glucagon-like peptide-1. Nutrients are thought to stimulate insulin secretion through metabolic secretion coupling to generate metabolic signals, i.e. second messengers or coupling factors. Extensive research has been conducted to determine how nutrients are metabolized to generate these coupling factors, e.g. ATP and NADPH. Although leucine and ␣-ketoisocaproate (KIC) 3 are potent insulin secretagogues (1), the underlying mechanisms of leucine and KIC-stimulated insulin secretion (LSIS and KICSIS, respectively) remain elusive. A key question is whether their oxidative decarboxylation is required for induction of insulin secretion.
Early studies suggested that leucine and KIC catabolism appeared to be necessary for these nutrients to induce insulin secretion (2)(3)(4). Leucine catabolism begins with the transfer of its amino group to ␣-ketoglutarate (␣-KG), resulting in the formation of KIC and glutamate. The mitochondrial branchedchained aminotransferase (BCATm), which is expressed in most peripheral tissues except liver, catalyzes this readily reversible reaction. Consistent with a role of leucine/KIC catabolism in LSIS and KICSIS, isolated islets cells exhibited high catabolic rates of leucine and KIC (2,5). It was postulated that like glucose oxidation, KIC oxidation produced elevated ratios of ATP/ADP, NADH/NAD ϩ , and NADPH/NADP ϩ to signal insulin release (6). However, it was later found that KIC was not a very effective substrate for ATP production, and this led to a proposal that, in addition to generation of ␣-KG, metabolites of leucine/KIC such as acetoacetate may induce the insulinotropic effect of KIC (7,8).
Another mechanism involves the discovery in 1980, that leucine is an allosteric activator of glutamate dehydrogenase (GDH), the enzyme that oxidizes glutamate to ␣-KG with concomitant production of NADPH, a potential intracellular signal for insulin secretion (9). Oxidation of glutamine and glutamate by leucine-activated GDH induces insulin secretion (9 -11). The importance of this pathway is highlighted by the discovery of a dominant form of congenital hyperinsulinism associated with gain of function mutations of GDH that lead to its impaired sensitivity to inhibition by GTP and ATP (12,13). Patients with inherited GDH mutations exhibit hypoglycemia and hyperinsulinism, especially after high protein meals. Glu-tamine itself does not induce insulin release but markedly enhances insulin secretion evoked by either leucine or a nonmetabolized analogue, 2-aminobicyclo- [2,2,1]heptane-2-carboxylate, indicating one of the potential mechanisms of LSIS is leucine activation of GDH. MacDonald et al. (14,15) proposed that short chain acyl-CoAs produced by leucine catabolism as well as ␣-KG derived from leucine-activated glutamate oxidation were necessary for enhancing anaplerosis and insulin secretion. Additionally, Fahien and MacDonald (16) posited that by interacting with succinate, mevalonate produced from leucine catabolism could be a signal for insulin release. Thus, both leucine catabolism and activation of GDH could be required for leucine/KIC to stimulate insulin secretion.
However, the catabolism hypothesis has been challenged by several studies. Gao et al. (17) found that an aminotransferase inhibitor, amino-oxyacetic acid, abolished KICSIS but potentiated LSIS, suggesting a mechanistic difference in LSIS and KIC-SIS, i.e. activating GDH is more important for LSIS, and transamination may play an important role in KICSIS (17)(18)(19). Rabaglia et al. (20) also reported that KICSIS was blocked by the BCAT inhibitors, methyl-leucine or aminooxyacetic acid, implicating a mechanistic role for ␣-KG. These studies address the possible importance of ␣-KG formation via BCATm and perhaps GDH for KICSIS. In fact, the same group who proposed the catabolism mechanism later reported that intramitochondrial ␣-KG generation may regulate the insulin secretory potency of leucine and KIC (21).
A potential caveat for the above studies is that none of the inhibitors used are specific. Indeed, aminooxyacetic acid inhibits insulin secretion stimulated by glucose and other nutrients (18,22). All of the available inhibitors of BCATm preferentially block other transaminases (23)(24)(25). In this study, we used BCATm Ϫ/Ϫ mice to determine the precise role of this enzyme in KICSIS and LSIS, therefore resolving the catabolism versus transamination dispute that has lasted for more than 30 years. The biochemical mechanisms and potential role of KICSIS and LSIS in obesity and type 2 diabetes are further discussed.

MATERIALS AND METHODS
Animals-All animal experiments were approved by the Institutional Animal Care and Use Committee at the Pennsylvania State University College of Medicine. BCATm Ϫ/Ϫ mice were generated and maintained as described previously (26). As elevated plasma branched-chain amino acid concentrations in BCATm Ϫ/Ϫ mice might alter CNS neurotransmitters, all mice were offered a choice of two diets, rodent normal chow (Harland Teklad 2018, Madison, Wisconsin) that has 18% protein as a percent of total weight and a defined amino acid branchedchain amino acid-free diet (Dyets Inc., Bethlehem, PA) that has 17% amino acids as percent of total weight.
Islet Isolation and Evaluations of Insulin Secretion Using Static Incubation and Islet Perifusion-Intact islets were isolated from mice using collagenase (Sigma-Aldrich) digestion and Histopaque 1077 (Sigma-Aldrich) gradient centrifugation as described previously (27). Islets were washed twice with RPMI 1640 cell culture medium supplemented with 10 mM glucose/2 mM glutamine, 10% fetal bovine serum, and 1% antibiotics-antimycotics. Batches of ϳ200 islets were cultured over-night in 6-cm dishes in a 37°C incubator in an atmosphere of humidified air and 5% CO 2 .
For insulin secretion experiments using static incubation, islets were preincubated in a CO 2 incubator for 60 min with 6 ml of Krebs-Ringer bicarbonate-HEPES buffer (KRBH) containing 0.5% BSA and gassed with 95% O 2 and 5% CO 2 . The KRBH contained 4.7 mM KCl, 2.5 mM CaCl 2 , 1.2 mM KH 2 PO 4 , 1.2 mM MgSO 4 , 118 mM NaCl, 29 mM NaHCO 3 , and 10 mM HEPES. After preincubation, batches of five size-matched islets were pipetted into a 24-well Petri dish and treated with 2 ml of KRBH buffer containing 16.7 mM glucose, 10 mM leucine/2 mM glutamine, 10 mM KIC/2 mM glutamine, or 10 mM dimethyl ␣-KG/2 mM glutamine in the 37°C incubator for 60 min. To stop insulin secretion, the Petri dish was placed on ice, and 1 ml of the supernatant solution was collected for insulin assay using an ELISA kit.
Islet perifusion experiments were performed as described previously (28). Briefly, islets were cultured for 3 days in RPMI 1640 containing 10 mM glucose/2 mM glutamine. Then 100 islets were "hand-picked" under a light microscope and loaded into a perifusion chamber and immediately perifused with KRBH buffer plus 0.25% BSA using a Bio-Rad Econo Gradient pump. The perfusates and perifusion chamber were maintained at 37°C and continuously gassed with 95% O 2 and 5% CO 2 . Islets were perifused with 2 mM glutamine from 0 -120 min and a ramp of KIC (0 -25 mM) or leucine (0 -25 mM) from 50 -100 min and 30 mM KCl from 120 -130 min. The flow rate was 1 ml/min, and samples were collected every minute for insulin measurement.
Evaluation of Insulin Secretion Using in Situ Pancreas Perfusion-Mouse pancreas perfusion was performed as reported previously (29). Briefly, mice were anesthetized by intraperitoneal injection of pentobarbital (Nembutal, 70 mg/kg). A middle abdominal incision was performed to expose the portal vein and aorta, followed by ligations of superior mesenteric artery, splenic artery, and right renal artery. A 27-gauge needle was immediately inserted into the aorta and moved toward the celiac trunk. Silicon tubing (0.51 mm inner diameter/0.94 mm outer diameter) was then inserted into the portal vein. Hepatic arteries just below the liver and left renal artery were tied. The mouse was immediately sacrificed by cutting the diaphragm and heart. The mouse, perfusion buffer, and pump were then maintained in a temperature (37°C) and humidity-controlled apparatus made by the Vanderbilt University Apparatus Shop. The mouse was covered with a piece of plastic wrap to prevent desiccation of the abdominal organs. The celiac trunk was perfused with oxygenated KRBH containing 1% BSA and 3% Dextran T70 at 1 ml/min through the aortic catheter. The effluent was collected for insulin assay at 1-min intervals from a cannula placed in the portal vein. For KICSIS, the perfusion lasted 90 min and consisted of a 20-min washout period with 2. Biochemical Analyses-To measure islet contents of malate and ␣-KG, ϳ200 islets, which were cultured separately overnight for each treatment, were preincubated in KRBH/0.5% BSA for 60 min and then incubated in a 1.5 ml microfuge tube with KRBH mixed with either 16.7 mM glucose, 10 mM KIC/2 mM glutamine, or basal buffer for 60 min. The incubation buffer was quickly pipetted out, and the islets were washed once with cold KRBH without BSA and centrifuged at 100 ϫ g for 15 s. After removing the buffer, the islets were homogenized in 80 l of 0.6 M perchloric acid and centrifuged at 16,000 ϫ g for 5 min, and the supernatant was neutralized to ϳpH 7 with 20 l of 2 M K 2 CO 3 . Malate concentration was measured enzymatically using a fluorometric assay (30). The ␣-KG concentration was measured using a more sensitive indirect fluorometric assay (30). These values were normalized to islet pellet protein concentrations.
To measure islet ATP concentrations, ϳ100 overnight-cultured islets were preincubated in KRBH/0.5% BSA for 60 min, then five islets for each treatment were pipetted into a 1.5-ml microfuge tube and incubated with 0.6 ml of KRBH/0.5% BSA with or without a stimulator (either 16.7 mM glucose, 10 mM leucine/2 mM glutamine, 10 mM KIC/2 mM glutamine, 10 mM dimethyl ␣-KG/2 mM glutamine) at 37°C for 60 min. After pipetting 0.2 ml of buffer out for insulin assay, 0.4 ml of ice-cold 0.3 M perchloric acid was added to each tube, and the islets were briefly sonicated and centrifuged at 16,000 ϫ g for 5 min. A 0.6-ml aliquot of supernatant was reserved and neutralized to ϳpH 7 with 60 l of 3N KOH/25 mM MOPS. The supernatant was further diluted, and ATP concentration was measured using a Bioluminescent Assay kit (Sigma Aldrich).
To measure islet NADPH:NADP ratio, ϳ200 overnight-cultured islets per test sample, were preincubated in KRBH/0.5% BSA for 60 min and incubated in a 1.5-ml microfuge tube with KRBH mixed with either 10 mM KIC/2 mM glutamine or basal buffer for 60 min. The incubation buffer was then quickly pipetted out, and the islets were washed once with cold KRBH without BSA and centrifuged at 100 ϫ g for 15 s. After removing the buffer, the islets were homogenized in 90 l of buffer containing 10 mM Tris, pH 8.1, 1% Nonidet P-40, 0.5 mM PMSF, and 0.5% Triton X-100 and centrifuged at 16,000 ϫ g for 5 min. Forty microliters of supernatant were combined with 5 l of 2 mM NaOH to destroy NADP ϩ (for NADPH assay) and 5 l of 2 mM HCl to destroy NADPH (for NADP ϩ assay). The samples were then neutralized with 5 l of 2 mM HCl and 5 l of 2 mM NaOH, respectively. NADPH and NADP ϩ in islet extracts were measured using an enzymatic cycling assay as described previously in detail (30,31). NADPH and NADP ϩ standard curves were run in parallel with the samples to calculate NADPH: NADP ϩ ratios in islet samples.
To measure pancreatic insulin content, the whole pancreas was surgically removed, weighed, and homogenized with an extract solution containing 75% ethanol/1.5% acetate and further extracted overnight at 4°C and diluted with phosphatebuffered saline for insulin assay using an ELISA kit (ALPCO). Insulin concentrations in samples from islet perifusion and pancreas perfusion were measured using a radioimmunoassay kit (Linco).
Leucine, KIC, and Glucose Oxidation-Batches of ϳ200 overnight-cultured islets were preincubated in KRBH/0.5% BSA in a CO 2 incubator for 60 min. Groups of 25 islets were then placed into microfuge tubes housed within a sealed 20-ml scintillation vial, which was filled with 95% O 2 and 5% CO 2 . Incubations were performed at 37°C for 2 h in 60 l of KRBH buffer containing 10 mM KIC/2 mM glutamine (with 1 Ci/ml of [U-14 C]KIC), 10 mM leucine/2 mM glutamine (with 1 Ci/ml of [U-14 C]leucine) or 16.7 mM glucose (with 1 Ci/ml of [U-14 C]glucose). After incubation, 300 l of phenethylaminemethanol was injected onto the filter paper placed in the central well to trap CO 2 released from incubation buffer by injecting 20 l of imididazole/HCl buffer (200 mM, pH 6). Disintegrations per minute in the filter papers was counted, and substrate oxidation was calculated from radioactivity in 14 CO 2 divided by total 14 C-labeled radioactivity added to islets. Separate blanks without islets were included in each experiment to correct nonspecific 14 CO 2 production. Meanwhile, the conversion of KIC into amino acids was evaluated in the same experiment by isolating radioactive amino acids in homogenized incubated islets using ion-exchange columns (Dowex 50Wx8 -200) as described (2). [U-14 C]leucine and [U-14 C]glucose were purchased from PerkinElmer Life Sciences. [U-14 C]KIC was enzymatically prepared from [U-14 C]leucine as described previously (32), and the KIC purity was examined by ascending TLC chromatography in a butan-1-ol/methanol/water (2:2:1) solvent system and visualized using a GE Typhoon instrument.
Islet Morphometry-The proportion of islet ␤-cell surface area versus surface area of the whole pancreas was determined planimetrically as detailed previously (33). To obtain information on potential changes in endocrine cell cluster/islet size dynamics, the relative surface area and number of ␤-cell clusters and very small islets (Ͼ400 m 2 or Ͼ23 m average diameter) were tallied for each animal using the same sections used for ␤-cell mass measurements as described (33). Image files were analyzed using NIH ImageJ.
Statistical Analysis-A two-tailed nonpaired t test was used to assess the difference between two groups. One-way analysis of variance with Newman-Keuls post tests was used to analyze data obtained from more than two groups or multiple time points. Values are mean Ϯ S.E., and p values Ͻ 0.05 were considered significantly different.

Effects of BCATm Disruption on Nutrient-stimulated Insulin
Secretion-We first measured insulin secretion using static incubation. As shown in Fig. 1A, glucose at 16.7 mM and leucine/glutamine at 10/2 mM induced 7-8-and 5-6-fold increases of insulin secretion, respectively, when compared with the nontreated group. These effects did not differ between BCATm Ϫ/Ϫ and BCATm ϩ/ϩ islets. KIC/glutamine at 10/2 mM induced a 9-fold increase in insulin secretion in BCATm ϩ/ϩ islets; however, it did not induce any insulin secretion in BCATm Ϫ/Ϫ islets, suggesting that loss of KIC/glutamate transamination to ␣-KG and leucine abolished KICSIS. The TCA intermediate and product of KIC/glutamate transamination, ␣-KG at 10 mM, caused a ϳ7-fold increase in insulin secretion in islets of both genotypes, consistent with a previous report that ␣-KG potently stimulates insulin secretion (20). While glutamine by itself does not induce insulin secretion, it has been shown to be readily converted to glutamate by a phosphate-dependent glutaminase in islets, and it is often added to LSIS experiments (28). Thus, glutamine was also used as a surrogate for glutamate in our experiments.
To characterize KICSIS further, we used islet perifusion and in situ pancreas perfusion techniques to measure KICSIS in BCATm Ϫ/Ϫ and BCATm ϩ/ϩ mice. In response to a ramp of KIC (from 0 to 25 mM in the presence of 2 mM glutamine), KIC induced robust dose-dependent insulin secretion in perifused BCATm ϩ/ϩ islets ( Fig. 2A). However, consistent with static incubation studies, KIC failed to stimulate insulin secretion in BCATm Ϫ/Ϫ islets. As a positive control, 30 mM KCl stimulated FIGURE 1. Effects of BCATm disruption on nutrient-stimulated insulin secretion. A, islets isolated from BCATm Ϫ/Ϫ and BCATm ϩ/ϩ mice were cultured overnight in PRMI 1640 supplemented with 10 mM glucose/2 mM glutamine. After preincubation in basal KRBH for 60 min, size-matched islets were pipetted into a 24-well Petri dish for a 60-min incubation with 2 ml of KRBH buffer containing 16.7 mM glucose, 10 mM leucine/2 mM glutamine, 10 mM KIC/2 mM Gln, and 10 mM dimethyl ␣-KG (DM␣-KG)/2 mM Gln within a 37°C 5% CO 2 incubator. Subsequently, the incubation buffer was collected for insulin immunoassay (n ϭ 8 from two independent experiments). B, a second study was performed using the same condition as in A except stimulating with branched-chain keto acids, 10 mM KIC, 20 mM KMV, 10 mM KC, and 10 mM KIV all in the presence of 2 mM Gln (n ϭ 8 -10 from two independent experiments). Values are mean Ϯ S.E.; a indicates p Ͻ 0.05 versus nontreated group of the same genotype; b indicates p Ͻ 0.05 versus BCATm ϩ/ϩ mice of the same treatment, and c indicates p Ͻ 0.05 versus BCATm ϩ/ϩ mice of KIC treatment.

FIGURE 2. Effects of BCATm disruption on KICSIS during islet perifusion and pancreas perfusion.
A, islets isolated from BCATm Ϫ/Ϫ and BCATm ϩ/ϩ mice were cultured for 3 days in RPMI 1640with 10 mM glucose/2 mM glutamine. 100 islets were loaded into a perifusion chamber and immediately perifused with KRBH buffer added with 0.25% BSA. Islets were perifused with 2 mM glutamine from 0 -120 min and a ramp of KIC from 0 -25 mM at 50 -100 min and 30 mM KCl at 120 -130 min. Samples were collected every minute for insulin assay. Values are mean Ϯ S.E. (n ϭ 3). B, after a surgical procedure as described under "Materials and Methods," the pancreas was perfused with oxygenated KRBH containing 1% BSA and 3% Dextran T70 through a catheter placed in the aorta. The effluent was collected at 1-min intervals from a cannula placed in the portal vein for insulin assay. insulin secretion to the same extent in both BCATm Ϫ/Ϫ and BCATm ϩ/ϩ islets, suggesting that the K ATP channel activity in BCATm Ϫ/Ϫ islets was unperturbed. During pancreas perfusion in BCATm ϩ/ϩ islets (Fig. 2B), KIC at 10 mM led to a brief but strong first phase of insulin secretion followed by an even greater second phase of insulin secretion, and a KIC/glutamine mixture caused a modest increase in insulin secretion compared with KIC alone (a 19% increase in area under curve for KIC versus KIC/glutamine: 92.7 Ϯ 9.2 versus 110.4 Ϯ 12.6, p Ͻ 0.05, n ϭ 3-4). In contrast, neither the mixture nor KIC alone caused appreciable insulin secretion in BCATm Ϫ/Ϫ pancreas. Because KCl-stimulated insulin secretion was similar in both types of mice, it is unlikely that there is a general problem with our methodology of insulin secretion measurement. Therefore, these results suggest that BCATm-catalyzed conversion of KIC to leucine and the consequent formation of ␣-KG from glutamate are essential for KICSIS. The very small level of KICSIS during pancreas perfusion in BCATm Ϫ/Ϫ mice (Fig. 2B) could simply result from leucine and ␣-KG converted from KIC/glutamate by nonspecific transaminases in the intact pancreas.
During pancreas perfusion, insulin secretion responses to glucose, arginine, or glucose/arginine were similar between BCATm Ϫ/Ϫ and BCATm ϩ/ϩ mice (Fig. 3A). The stimulatory patterns of these secretagogues during perfusion were typical. Arginine stimulates insulin release by directly affecting the K ATP channel and membrane potential, and the additive effect of glucose and arginine on insulin secretion has been reported (35). These data further support the conclusion that the loss of leucine catabolism in BCATm Ϫ/Ϫ mice does not affect other secretagogues.
We also found that LSIS measured using islet perifusion appeared normal in isolated islets from BCATm Ϫ/Ϫ mice when compared with BCATm ϩ/ϩ mice (Fig. 3B). Notably, LSIS in response to increasing concentrations of leucine is not dose-dependent. Leucine at low concentration led to a strong but brief first phase of insulin secretion. At higher concentrations, it induced a weaker second phase of insulin secretion. Also, LSIS is much less potent than KICSIS in BCATm ϩ/ϩ islets ( Fig. 2A  and 3B). These data suggest that disruption of BCATm in mice does not affect LSIS in isolated mouse islets. In response to 30 mM KCl, the magnitude of insulin release in this experiment was comparable to those seen in the KIC experiment (data not shown), again validating the islet perifusion protocol.

Effects of BCATm Disruption on KIC, Leucine, and Glucose Oxidation and Transamination of KIC to Amino Acids in Islets-
To confirm that BCATm disruption truly blocks leucine and KIC transamination in isolated islets, we used 14 C-labeled KIC and leucine to measure substrate oxidation and transamination. As expected, loss of BCATm did not affect KIC oxidative decarboxylation (Fig. 4A); however, it totally blocked the conversion of KIC/glutamine to amino acids (presumably leucine) (Fig. 4B). Notably, in BCATm ϩ/ϩ islets, the rate of KIC conversion to amino acids was 3.7-fold higher than that of KIC oxidation. BCATm disruption also abolished leucine oxidative decarboxylation (Fig. 4C). The small residual leucine oxidation in BCATm Ϫ/Ϫ islets could result from nonspecific transamination and/or high sensitivity of the isotope method. Consistent with unaltered glucose-stimulated insulin secretion, glucose oxidation did not differ between BCATm ϩ/ϩ and BCATm Ϫ/Ϫ islets (Fig. 4D).
Effects of BCATm Disruption on Islet Concentrations of ATP, ␣-KG, and Malate as well as NADPH/NADP ϩ Ratio-Insulin secretion stimulated by nutrients is strongly associated with mitochondrial metabolism and ATP production. Therefore, we measured islet concentrations of ATP and key TCA cycle intermediates as well as the NADPH/NADP ϩ ratio (Fig. 5). Compared with the respective nontreated group, ATP concentrations were significantly increased in cultured islets treated with glucose, leucine/glutamine, and ␣-KG/glutamine from both BCATm ϩ/ϩ and BCATm Ϫ/Ϫ mice. ATP levels were similar under these different conditions regardless of the BCATm sta- tus (Fig. 5A). ATP concentrations were significantly elevated in BCATm ϩ/ϩ but not BCATm Ϫ/Ϫ islets after KIC/glutamine stimulation. Thus, nutrients consistently stimulated ATP production and insulin secretion in both BCATm ϩ/ϩ and BCATm Ϫ/Ϫ islets with the exception that KIC specifically failed to do so in islets lacking BCATm.
NADPH is emerging as one of the strongest candidate coupling factors for glucose-stimulated insulin secretion (36). As shown in Fig. 5D, compared with the nontreated group, 10 mM KIC/2 mM glutamine treatment caused an ϳ2-fold increase in the NADPH/ NADP ϩ ratio in BCATm ϩ/ϩ but not BCATm Ϫ/Ϫ islets.
Effects of BCATm Disruption on Pancreatic Insulin Content, ␤-Cell Mass, and Morphology-Because leucine is a potent stimulator of the mTOR signaling pathway, which regulates protein synthesis and cell size (37), we measured whether loss of leucine metabolism influences pancreatic insulin content, ␤-cell mass, and islet morphology (supplemental Fig. 1). Neither pancreatic insulin content nor ␤-cell mass differed between BCATm ϩ/ϩ and BCATm Ϫ/Ϫ mice. No alterations in islet architecture were detected between the groups of mice, and no differences were observed in the relative size and num-bers of islets (data not shown). We have also found that leucinebut not insulin-stimulated mTOR activation was largely abolished in isolated adipocytes, perfused liver and heart as well as cultured primary fibroblasts from BCATm Ϫ/Ϫ mice. 4 It is highly likely the same is true in ␤-cells of these animals; thus, it would be complex to determine how altered mTOR signaling regulates insulin synthesis and ␤-cell development in these mice. Nonetheless, the apparent normal ␤-cell morphology was consistent with unaltered insulin secretion by most tested secretagogues except BCKAs in BCATm Ϫ/Ϫ mice.
Effects of BCATm Disruption on Plasma Glucose and Insulin Responses to Fasted Refeeding-Finally, to determine whether BCATm disruption affects insulin secretion in vivo, we measured plasma insulin and glucose concentrations in response to refeeding after a 21-h fast (Fig. 6). Body weight-corrected food intake during the 6-h refeeding period did not differ between 4 P. She and C. J. Lynch, unpublished data.  . Effects of BCATm disruption on islet contents of ATP, malate, and ␣-KG as well as the NADP/NAD ratio. A, After an overnight culture, five islets per condition were transferred into a 1.5-ml microfuge tube and incubated with 0.6 ml KRBH/0.5% BSA without or with stimulators (16.7 mM glucose, 10 mM leucine/2 mM glutamine), 10 mM KIC/2 mM Gln, or 10 mM dimethyl ␣-KG (DM␣-KG)/2 mM Gln at 37°C for 60 min. ATP was measured using a Bioluminescent Assay kit as described under "Materials and Methods." B and C, ϳ200 overnight-cultured islets for each treatment were preincubated in KRBH/0.5% BSA for 60 min and then incubated in a 1.5-ml microfuge tube stimulated with 16.7 mM glucose, 10 mM KIC/2 mM Gln, or basal buffer for 60 min. Malate was measured using an enzymatic analysis coupled with fluorometry, and ␣-KG was measured using more sensitive indirect fluorometry. D, ϳ200 overnight-cultured islets for each treatment were preincubated in KRBH/0.5% BSA for 60 min and then incubated in a 1.5-ml microfuge tube stimulated with 10 mM KIC/2 mM Gln or basal buffer for 60 min. NADPH/ NADP ϩ ratios were measured using an enzymatic cycling assay coupled with fluorometry. Values are mean Ϯ S.E. a, p Ͻ 0.05 versus nontreated group of the same genotype; b, p Ͻ 0.05 versus BCATm ϩ/ϩ mice of the same treatment (n ϭ 10 from two independent experiments for ATP assay; n ϭ 4 -5 for malate and ␣-KG assays; and n ϭ 5 for the NADPH/NADP ϩ ratio measurement. female BCATm ϩ/ϩ and BCATm Ϫ/Ϫ mice. Compared with BCATm ϩ/ϩ mice, whereas basal fasting plasma glucose and insulin concentrations did not differ, plasma insulin concentrations at 3, 4.5, and 6 h of refeeding were decreased more than by half in BCATm Ϫ/Ϫ mice, and plasma glucose concentration at 1 h of refeeding was markedly lowered in these animals. Additionally, after 3 h of refeeding, glycogen contents in liver (data not shown) and skeletal muscle (38) were higher in BCATm Ϫ/Ϫ than BCATm ϩ/ϩ mice. These data suggest improved insulin sensitivity and glucose tolerance in mice without BCATm as reported previously (26) and much lower insulin secretion in response to refeeding in these mice.

DISCUSSION
In this study, we have found that disruption of the first step of branched-chain amino acid metabolism specifically abolished insulin secretion stimulated by BCKAs without affecting insulin secretion by glucose, arginine, glucose/arginine, leucine/ glutamine, ␣-KG/glutamine, or KCl. The specific loss of BCKAstimulated insulin secretion in islet cells lacking BCATm indicates that transamination of BCKAs and glutamine to branched-chain amino acids and ␣-KG is essential for BCKAs to stimulate insulin secretion and that BCKA oxidation alone is not sufficient to induce insulin secretion. Furthermore, unaltered LSIS in BCATm Ϫ/Ϫ islets suggests that leucine transamination to KIC and its further catabolism is not necessary for LSIS.
Our study is consistent with previous studies in which KIC-SIS was found to be blunted by transaminase inhibitors (17,18,20) and strongly supports the notion that ␣-KG generation by BCATm-catalyzed transamination of KIC/glutamate and GDH-catalyzed glutamate oxidation, which is activated by leucine, is important for KICSIS (17,20,21). It has been reported that KIC and KC, the transamination products of leucine and norleucine, respectively, stimulate insulin secretion with equal potency, whereas KMV, the transamination product of isoleucine, is a much weaker secretagogue, and KIV, the transamination product of valine, is not a secretagogue (4,9,34). Although ␣-KG can be produced through the reverse transamination of all of these BCKAs and glutamate, the discrepancy in stimulatory potency of these BCKAs has been explained to depend on whether the corresponding amino acid is a activator of GDH and the transamination capacity of this amino acid with ␣-KG to reduce ␣-KG production (17,21). Leucine is a potent GDH activator, whereas isoleucine can weakly activate GDH, and valine cannot activate GDH; therefore, the transamination-mediated loss of ␣-KG is differentially compensated by GDH-catalyzed glutamate oxidation to ␣-KG (21). KC and glutamate are efficiently transaminated to norleucine and ␣-KG, whereas the transamination of norleucine and ␣-KG to KC and glutamate is very low in islets, avoiding the loss of ␣-KG produced from KC transamination (8). In islets lacking BCATm, ␣-KG cannot be generated from the transamination of BCKAs and glutamate, in parallel with the loss of insulinotropic effect of all BCKAs.
It has been proposed that ␣-KG directly stimulates insulin secretion (20), and we confirmed that ␣-KG stimulated insulin secretion in both BCATm ϩ/ϩ and BCATm Ϫ/Ϫ islets. After KIC stimulation, islet ␣-KG concentration was lower in BCATm ϩ/ϩ than BCATm Ϫ/Ϫ mice, whereas KICSIS occurred only in BCATm ϩ/ϩ islets. This result argues against ␣-KG being a direct secretory coupling factor. Despite the lowered ␣-KG content in BCATm ϩ/ϩ islets stimulated with KIC, we postulate that ␣-KG generated from KIC transamination and glutamate oxidation enhances anaplerosis and TCA cycle activity to generate coupling factors of insulin secretion. Enhanced anaplerosis not only promotes mitochondrial metabolism but also increases certain mitochondrial products that may act as second messengers to amplify nutrient-stimulated insulin secretion (15,39,40). Our finding that in response to KIC stimulation, islet ATP concentration and NADPH/NADP ϩ ratio were elevated in BCATm ϩ/ϩ but not BCATm Ϫ/Ϫ islets is consistent with this hypothesis. NADPH was found to stimulate insulin granule exocytosis (41). The NADPH/NADP ϩ ratio was found FIGURE 6. Effects of BCATm disruption on food intake, plasma insulin, and glucose concentrations in response to fasted refeeding. Female mice were fasted for 21 h and then refed with diet choice of normal chow/amino acid purified branched-chain amino acid-free diet for 6 h. Body weight (B.W.)normalized food intake during the 6-h refeeding period was recorded (A), and blood samples were collected at 0, 1, 3, 4.5, and 6 h of refeeding for measuring plasma insulin (B) and glucose (C). Values are mean Ϯ S.E. *, p Ͻ 0.05 versus BCATm ϩ/ϩ mice at the same time point; n ϭ 13 for BCATm ϩ/ϩ and 12 for BCATm Ϫ/Ϫ mice.
Collectively, our data suggest that KICSIS and LSIS occur by enhancing the metabolism of glutamine/glutamate to ␣-ketoglutarate, which, in turn, is metabolized to produce the intracellular signals such as ATP and NADPH for insulin secretion. This finding raises a reasonable question, based on the fact that plasma KIC concentration is in the ten micromolar range. Is KICSIS really physiologically relevant? We think that this is the case for the following reasons. Basal plasma leucine concentrations are at least 10-fold higher than that of KIC, and leucine was shown to rise severalfold over baseline after a meal in rats (46). Although islets possess BCATm, it should be noted that the pancreatic acinar tissue contains the highest amount of BCATm in the body, which is ϳ5-fold higher than that in islets (47). Thus, it is very likely that much of the KIC in islets could actually be diffused from surrounding acinar tissue where KIC is transaminated from leucine. As a result, in islets, a considerably higher amount of intramitochondrial ␣-KG could be generated from KIC and glutamate transamination, as compared with leucine activation of islet GDH alone. Leucine thus could augment its insulinotropic effect through this paracrine control of exocrine pancreas over ␤-cell function. Further supporting relevance is the observation that KIC causes hypersecretion of insulin in diabetes-susceptible BTBR mice and high fat diet-fed mice (20,48). It has also been reported that the prevalence of insulin hypersecretion is higher than that of insulin resistance in obese subjects (49) and in some instances may represent a primary metabolic defect (50). Thus, further studies are warranted to investigate whether insulin hypersecretion stimulated specifically by KIC/leucine/glutamine/glutamate plays a general role in the pathogenesis of obesity-associated diabetes.