Glutathionylation State of Uncoupling Protein-2 and the Control of Glucose-stimulated Insulin Secretion*

Background: Proton leak through UCP2 is modulated by glutathionylation, and UCP2 modulates GSIS. Results: Glutathionylation of UCP2 amplifies GSIS from β cells. Matrix ROS activates UCP2-desensitizing GSIS. Conclusion: Reversible glutathionylation of UCP2 aids in regulating GSIS. Significance: Findings enhance our understanding of the role of redox circuits in the modulation of GSIS. The role of reactive oxygen species (ROS) in glucose-stimulated insulin release remains controversial because ROS have been shown to both amplify and impede insulin release. In regard to preventing insulin release, ROS activates uncoupling protein-2 (UCP2), a mitochondrial inner membrane protein that negatively regulates glucose-stimulated insulin secretion (GSIS) by uncoupling oxidative phosphorylation. With our recent discovery that the UCP2-mediated proton leak is modulated by reversible glutathionylation, a process responsive to small changes in ROS levels, we resolved to determine whether glutathionylation is required for UCP2 regulation of GSIS. Using Min6 cells and pancreatic islets, we demonstrate that induction of glutathionylation not only deactivates UCP2-mediated proton leak but also enhances GSIS. Conversely, an increase in mitochondrial matrix ROS was found to deglutathionylate and activate UCP2 leak and impede GSIS. Glucose metabolism also decreased the total amount of cellular glutathionylated proteins and increased the cellular glutathione redox ratio (GSH/GSSG). Intriguingly, the provision of extracellular ROS (H2O2, 10 μm) amplified GSIS and also activated UCP2. Collectively, our findings indicate that the glutathionylation status of UCP2 contributes to the regulation of GSIS, and different cellular sites and inducers of ROS can have opposing effects on GSIS, perhaps explaining some of the controversy surrounding the role of ROS in GSIS.

The role of reactive oxygen species (ROS) in glucose-stimulated insulin release remains controversial because ROS have been shown to both amplify and impede insulin release. In regard to preventing insulin release, ROS activates uncoupling protein-2 (UCP2), a mitochondrial inner membrane protein that negatively regulates glucose-stimulated insulin secretion (GSIS) by uncoupling oxidative phosphorylation. With our recent discovery that the UCP2-mediated proton leak is modulated by reversible glutathionylation, a process responsive to small changes in ROS levels, we resolved to determine whether glutathionylation is required for UCP2 regulation of GSIS. Using Min6 cells and pancreatic islets, we demonstrate that induction of glutathionylation not only deactivates UCP2-mediated proton leak but also enhances GSIS. Conversely, an increase in mitochondrial matrix ROS was found to deglutathionylate and activate UCP2 leak and impede GSIS. Glucose metabolism also decreased the total amount of cellular glutathionylated proteins and increased the cellular glutathione redox ratio (GSH/GSSG). Intriguingly, the provision of extracellular ROS (H 2 O 2 , 10 M) amplified GSIS and also activated UCP2. Collectively, our findings indicate that the glutathionylation status of UCP2 contributes to the regulation of GSIS, and different cellular sites and inducers of ROS can have opposing effects on GSIS, perhaps explaining some of the controversy surrounding the role of ROS in GSIS.
Blood glucose homeostasis is highly dependent on pancreatic ␤ cell mitochondrial energetics (1). When blood glucose levels rise, ␤ cells import and oxidize glucose to support ATP synthesis. Glucose oxidation is coupled to the formation of an electrochemical gradient across the mitochondrial inner membrane, referred to as protonmotive force (PMF), 4 which is then used to drive ATP production. The increase in cytoplasmic ATP/ADP deactivates K ATP channels prompting Ca 2ϩ uptake and subsequent release of insulin (2). The importance of coupled oxidative phosphorylation in insulin release can be illustrated by the use of chemical uncouplers that depolarize the PMF impeding ATP synthesis and glucose-stimulated insulin release (3). Hence, effective glucose-stimulated insulin secretion (GSIS) is reliant on efficient coupling of glucose catabolism to ATP synthesis. (4,5).
For over 15 years, researchers have been trying to assign physiological functions to mitochondrial uncoupling protein 2 (UCP2), which was discovered and named based on its sequence homology to UCP1, a protein in brown adipose tissue that, when active, dissipates the PMF preventing ATP production. However, unlike UCP1, inducible proton leak through UCP2 does not play a thermogenic function but rather a cell signaling role (6). UCP2 activity and expression are associated with negative regulation of insulin secretion. Knock-out of UCP2 expression or inhibition of UCP2 function with genipin increases GSIS from pancreatic islets, observations that have been replicated in INS-1E cells (7)(8)(9). The putative signaling function of UCP2 is thought to stem from its capacity to control mitochondrial ROS emission (reviewed in Ref. 10). For instance, using a cell-specific UCP2 deletion mouse model, Robson-Doucette et al. (9) demonstrated that ␤ cell UCP2 has little effect on mitochondrial ATP production, but it significantly contributes to the control of mitochondrial ROS production, which in turn regulates GSIS. In support of this, various reports have shown that exposing ␤ cells (either insulinoma cells or in pancreatic islets) to low amounts of superoxide (O 2 . , generated artificially with menadione) or H 2 O 2 stimulates insulin release (reviewed in Refs. 9,[11][12][13][14]. Furthermore, Leloup et al. (15) showed that the induction of ROS emission from the electron transport chain stimulates insulin release to the same degree as glucose-mediated ATP production. Glucose metabolism has also been shown to increase intracellular ROS levels in rat islets, Min6 (mouse ␤ cell line), and INS-1 832/13 cells (rat ␤ cell line), conditions associated with GSIS (9,12). In addition to the regulation of GSIS-amplifying ROS signals, ROS are also important regulators of UCP2 function itself (1). In a series of publications, Brand and co-workers (16,17) showed that proton leak through the uncoupling proteins is acutely controlled by ROS. As there is a non-Ohmic relationship between PMF and mitochondrial ROS production, even minor increases in uncoupling cause significant decreases in mitochondrial ROS emission when PMF is high (18,19). Recently, Affourtit et al. (8) showed that proton leak through UCP2 decreases GSIS by diminishing ROS production. UCP2 is well known to regulate mitochondrial ROS production in many tissues and cell types (reviewed in Ref. 20). However, as discussed above, ROS also activate GSIS. It is therefore paradoxical that mitochondrial ROS amplify GSIS and also activate UCP2, a negative regulator GSIS. One potential explanation is that the cellular location of ROS genesis is important in controlling GSIS.
Reversible glutathionylation involves the formation of a disulfide linkage between a protein thiol and glutathione. This post-translational modification is required to modulate protein function in response to fluctuations in cell redox state (21). Recently, our group showed that reversible glutathionylation is required to modulate proton leak through UCP2 and UCP3 but not UCP1 (6,22). Small nontoxic increases in ROS deglutathionylate UCP2-and UCP3-activating proton leak, thereby diminishing mitochondrial ROS emission through a negative feedback loop. Conversely, glutathionylation deactivates leak through these proteins. We have established that reversible glutathionylation of UCP2 and UCP3 is required to acutely control mitochondrial ROS production (23).
Using Min6 cells as a model system, we set out to determine whether reversible glutathionylation of UCP2 plays a signaling role during GSIS. Pharmacological induction of glutathionylation with diamide (100 M), a powerful glutathionylation catalyst, inhibited proton leak through UCP2 and increased GSIS. These observations were confirmed in pancreatic islets. Intriguingly, the treatment of cells with H 2 O 2 (10 M) had a dual effect amplifying GSIS yet activating proton leak through UCP2. Using paraquat, a superoxide-generating bipyridine that accumulates in mitochondria, we found that matrix ROS actually inhibits GSIS by activating the UCP2 leak. Hence, our results show that glutathionylation of UCP2 deactivates proton leak and amplifies GSIS. We also demonstrate that the impact of ROS on GSIS depends on the ROS location. The implications of ROS signaling in the matrix versus the cytoplasm are also discussed.

MATERIALS AND METHODS
Cell Culture and Treatment-Min6 insulinoma cells were routinely cultured in T75-cm 2 flasks on plastic and maintained in high glucose (25 mM) Dulbecco's modified Eagle's medium (DMEM; 4 mM glutamine, 1 mM pyruvate) containing 10% fetal bovine serum (FBS), 2% antibiotics/antimycotics, and 50 M ␤-mercaptoethanol. Medium was changed every 2 days, and cells were split every 4 days. For cell splitting, medium was aspirated, and the cell monolayer was treated with full strength trypsin (Invitrogen) for 1 min at 37°C. Trypsin was then deactivated with 3 volumes of medium, and cells were pelleted by centrifugation. The pellet was then resuspended in medium and split into new cultures. For experiments, cells were diluted to ϳ80,000 cells/ml in clear or black 96-well plates or in Seahorse tissue culture plates 2 days prior to experimentation. For all paraquat assays, cells were treated with paraquat (0 -500 M) for 18 h prior to experimentation. On the day of the experiments, medium was aspirated; cells were washed once with Krebs-Ringer buffer (KRB: 128 mM NaCl, 4.8 mM KCl, 1.2 mM KH 2 PO 4 , 1.2 mM MgSO 4 , 2.5 mM CaCl 2 , 10 mM HEPES, 0.1% (w/v) BSA, pH 7.4, 5 mM NaHCO 3 added fresh on the day of experiments) and incubated for 1 h at 37°C in KRB containing 1 mM glucose. Cells were then incubated for 1 h at 37°C in KRB containing 1 or 25 mM glucose and supplemented with either diamide (0 -1000 M), hydrogen peroxide (H 2 O 2 ; 0 -100 M or 5 mM as a control for ROS and cytotoxicity assays), or biotinylated glutathione ethyl ester (BioGEE, 1 mM). Following the incubation, the final assay medium was removed and stored at 4°C for insulin release determinations.
UCP2 Knockdown-Following 1 day of growth, cells were treated with Polybrene (2 g/ml, Santa Cruz Biotechnology) with either UCP2 shRNA (shUCP2; Santa Cruz Biotechnology) or scrambled (control, shCtl; Santa Cruz Biotechnology) shRNA lentiviral particles (5000 infectious viral particles/ml, Santa Cruz Biotechnology) for 48 h. Cultures were re-supplemented with fresh medium devoid of lentiviral particles or Polybrene and incubated for an additional 24 h in a medium, including puromycin (1 g/ml, Santa Cruz Biotechnology). Transduced cells were then lifted and treated accordingly for experimentation. The cell transduction protocol was performed according to the manufacturer's instructions and was optimized to minimize cell death and ensure maximum UCP2 knockdown.
Animals and Islet Isolation-The loxUCP2 mice were a gift from Dr. Bradford Lowell (24). ␤ cell-specific UCP2 deletion was accomplished by crossing loxUCP2 mice with rat insulin promoter-driven Cre recombinase (RIPCre) mice (The Jackson Laboratory, Bar Harbor, ME). Mice were genotyped using standard PCR of ear notch DNA. RIPCre mice were chosen as controls for experimentation because RIPCre and floxed mice (mice that express the floxed Ucp2 gene without Cre) gave similar results. All mice (10 -13 weeks old) were age-and sexmatched and maintained on a 129J-C57BL/6-mixed background. A mixture of male and female mice were used for experiments. All animal experiments were approved by the University of Toronto Animal Care Committee, and animals were handled according to the guidelines of the Canadian Council of Animal Care.
Islets were isolated from anesthetized mice by perfusing the pancreas via the pancreatic duct with a solution of collagenase type-V (0.8 mg/ml) (Sigma) in RPMI 1640 medium (11.1 mM glucose) supplemented with 1% v/v penicillin/streptomycin (Invitrogen) and 1% v/v L-glutamine (Invitrogen). The pancreas was digested for 17 min at 37°C before the addition of RPMI 1640 medium (11.1 mM glucose) supplemented with 10% v/v FBS, 1% v/v penicillin/streptomycin (Invitrogen), and 1% v/v L-glutamine (Invitrogen) to stop the digestion.
Insulin Release-Assay medium was collected from Min6 cell cultures that were incubated in the various conditions described above. Insulin levels were tested using the Cisbio HTRF insulin assay kit (CisBio Bioassays, Bedford, MA), as described by the manufacturer's instructions. Results were normalized to total cellular protein levels, which were determined using the Bradford assay.
Islet GSIS was tested as described previously (9) with a few modifications. Islets were washed with ice-cold KRB and preincubated at 37°C in 2.8 mmol/liter glucose (Ϯ10 M diamide) for 1 h. Tubes were put in ice water, and the 2.8 mmol/liter glucose buffer was replaced with ice-cold 2.8 or 16.7 mmol/liter glucose KRB (Ϯ10 M diamide) and incubated for 30 min at 37°C. All insulin values were corrected to DNA content. Data were represented as the fold change in insulin release relative to islets treated with 0 M diamide.
In Situ Monitoring of Cellular Bioenergetics-Characteristics of Min6 cellular bioenergetics were determined using the Seahorse extracellular flux analyzer (Seahorse Bioscience, North Billerica, MA). Cells were washed once with warmed PBS and then incubated for 30 min at 37°C and ambient CO 2 in HCO 3free DMEM containing 1 mM glucose, 4 mM glutamine, 1 mM pyruvate (pH 7.4). Diamide (100 M) or H 2 O 2 (10 M) was injected immediately followed by a 5-min incubation and the measurement of resting oxygen consumption rate (OCR). Cells were then treated sequentially with glucose (25 mM final concentration), oligomycin (0.2 g/ml), and antimycin A (1 M) to test the impact of diamide or H 2 O 2 on glucose-stimulated, proton leak-dependent, and extra-mitochondrial respiration, respectively. Mitochondrial respiration was calculated by subtracting extra-mitochondrial respiration values. Corrected OCR values were normalized to total protein/well using the Bradford assay. Changes in respiration rates relative to resting respiration values were determined by expressing the percent change in OCR relative to resting OCR.
ROS Measurements-Total cellular and matrix levels of ROS were measured using dichlorofluorescein diacetate (DCFHDA; 20 M) and MitoSOX (20 M), respectively. Reagent concentrations and durations of exposure were optimized prior to assays. Cells were loaded with DCFHDA or MitoSOX during glucose starvation (1 h of incubation in KRB containing 1 mM glucose), washed twice with KRB, and then incubated for 1 h in 1 or 25 mM glucose. Following the incubation, cells were washed twice with PBS. DCFHDA and MitoSOX fluorescence was measured at excitation/emission wavelengths of 480/530 nm and 514/585 nm. Cells not exposed to DCFHDA or MitoSOX were used to test background fluorescence. Results were normalized to protein content/well using Bradford assay and background fluorescence.
TMRE Fluorescence-Mitochondrial membrane potential was measured in situ using TMRE under nonquench mode conditions. Following cell starvation, cells were incubated for 1 h at 37°C in KRB containing 5 mM NaHCO 3 , 1 or 25 mM glucose, and diamide or H 2 O 2 and 10 nM TMRE. Incubation medium was then removed; cells were washed twice with PBS, and fluorescence was measured at excitation/emission wavelengths of 548/573 nm. Cells not incubated in TMRE were used to assess background fluorescence. Results were normalized to protein content/well using the Bradford assay and background fluorescence.
Cytotoxicity Assays-The effect of diamide on Min6 cell viability was tested using 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT; Calbiochem) and propidium iodide (PI; Sigma) assays (25). Min6 cells were starved of glucose and then incubated in KRB containing 5 mM NaHCO 3 , 25 mM glucose, diamide, and MTT assay reagent. Cells exposed to 5 mM H 2 O 2 served as the control. MTT assays were conducted as described in the manufacturer's instructions. For PI assays, following the 1 h of incubation of cells with diamide and 25 mM glucose, cells were washed once with PBS and then incubated for 10 min in PI diluted in PBS (10 g/ml). Cells were then washed twice, and plates were read at excitation/emission wavelengths of 530/615 nm. PI results were normalized to protein content as determined by the Bradford assay.
Glutathione Pool and ATP/ADP-For GSH/GSSG, GSH bound to protein, and ATP/ADP were measured using HPLC as described previously (20,25). For GSH and GSSG levels, cells were trypsinized, washed once with PBS, and then diluted to 0.5 mg/ml in ice-cold 1% (w/v) trifluoroacetic acid (TFA; 1% w/v), meta-phosphoric acid solution and incubated on ice for 10 min. Precipitate was removed by centrifugation for 10 min at 12,000 ϫ g at 4°C. The supernatant was collected and stored at Ϫ80°C. The protein pellet was treated with 1 N KOH for 20 min at room temperature; the base was neutralized with perchloric acid, and the precipitate was removed by centrifugation. The pellet was then discarded, and the supernatant was collected for analysis. On the day of the experiments, samples were injected into an Agilent 1100 Series HPLC equipped with a Pursuit C 18 column (150 ϫ 4.6 mm, 5 m, Agilent Technologies). For proper separation of GSH and GSSG, a flow rate of 1 ml/min was used with a mobile phase consisting of 0.1% (w/v) TFA and HPLC-grade methanol in a 90:10 ratio. GSH and GSSG were detected using an Agilent UV-visible variable wavelength detector operating at 215 nm. Retention times were confirmed by injecting GSH and GSSG standards. GSH and GSSG were quantified using Agilent Chemstation software.
For ATP and ADP level determinations, cells were collected, washed once with PBS, and then diluted to 0.5 mg/ml in a 0.5% (v/v) solution of perchloric acid and incubated on ice for 10 min. Following removal of the protein precipitate by centrifugation, the sample was injected into a C 18 hydrophilic reverse phase column operating at a flow rate of 0.7 ml/min (Synergy Hydro-RP; 4 m; 250 ϫ 4.6 mm, Phenomenex). The mobile phase was 20 mM KH 2 PO 4 , pH 2.9. ATP and ADP were detected at 254 nm for nucleotides. Quantification was performed by injecting varying amounts of ATP and ADP.
Western Blotting-Min6 cells were trypsinized, washed once with PBS, and lysed on ice in RIPA buffer. Protein was diluted in Laemmli buffer and then electrophoresed on SDS-polyacrylamide denaturing gels. Upon completion, gel slabs were removed and equilibrated in transfer buffer, and then proteins were electroblotted onto nitrocellulose membranes. Membranes were blocked for 1 h with 5% (w/v) nonfat skim milk and then probed overnight with anti-UCP2 antibodies (N-19, Santa Cruz Biotechnology), glutaredoxin-1 (Grx1, Abcam), or glutaredoxin-2 (Grx2, Abcam). Probing for succinate dehydrogenase (Santa Cruz Biotechnology) served as the loading control. For overnight incubation, antibodies were diluted in Tris-buffered saline, 2% (v/v) Tween 20. Bands were visualized by incubating membranes for 1 h with anti-goat or anti-mouse HRPconjugated secondary antibodies followed by a 5-min incubation in chemiluminescent substrate (ECL kit, Thermo Scientific).
UCP2 Glutathionylation Status-The glutathionylation status of UCP2 was determined using BioGEE (Invitrogen) and immunoblotting, as described previously (17). Min6 cells were pre-starved with KRB solution containing 1 mM glucose and then treated with 1 mM glucose or 25 mM glucose ϩ 1 mM BioGEE with or without 10 M H 2 O 2 for 1 h at 37°C. Cells were washed once with PBS and lysed in RIPA buffer containing protease inhibitors and 50 mM N-ethylmaleimide (to deactivate any unbound BioGEE and block unmodified thiol residues). Protein lysate was diluted to 2 mg/ml and incubated overnight under constant agitation in streptavidin beads at 4°C to elute proteins modified with BioGEE. A more deglutathionylated protein would bind more BioGEE. BioGEE-modified proteins were eluted by centrifugation (150 ϫ g for 5 min at 4°C); the supernatant was placed on ice, and the pellet was then treated with 4 M urea (in PBS, pH 7.4) to detach the BioGEEylated proteins from the streptavidin. The solution was centrifuged (150 ϫ g for 5 min at 4°C); the pellet discarded, and the resulting supernatant was placed on ice. Protein samples were then subjected to electrophoresis and then immunoblotted for the presence of UCP2 under reducing conditions, as described above. The amount of protein used in the initial elution (2 mg/ml) was loaded as an input control.
Immunoblot Detection of the Glutathionylated Proteome-Min6 cells treated with BioGEE were lysed and electrophoresed under nonreducing conditions. The amount of protein modified by BioGEE was detected by immunoblot using avidin-HRP antibody as described previously (17). Briefly, blocked membranes were washed twice with TBS-T and then incubated for 1 h in the dark in avidin-HRP antibody diluted in blocking solution (1:500, Abcam). Bands were visualized with chemiluminescent reagent (ECL kit, Thermo Scientific).
Statistical Analysis-Student's t tests were performed with Microsoft Excel software. One-way ANOVA with Fisher's protected least significant difference post hoc test or ANOVA repeated measures with Student's t tests were performed with StatView software. All values were expressed as means Ϯ S.E.

Low Doses of H 2 O 2 Amplify GSIS-
The role of ROS in enhancing GSIS is paradoxical in the sense that ROS can increase insulin release but also activate UCP2, which is known to negatively regulate GSIS. To confirm that low doses of ROS enhance GSIS, Min6 cells, a mouse insulinoma cell line that expresses UCP2 (supplemental Fig. 1 (25 mM) conditions. Although we did not observe an increase in GSIS when Min6 cells were exposed to 5 M H 2 O 2 , providing cells with 10 M did enhance GSIS (Fig. 1A), which is consistent with previous observations (12). Exposure of glucose-energized cells to 100 M H 2 O 2 increased GSIS further. However, cells exposed to 100 M H 2 O 2 also had high cellular ROS levels, indicating that the latter increase in extracellular insulin may be due to oxidative stress and cell damage (Fig. 1B). We also determined whether H 2 O 2 could stimulate insulin release under low glucose (1 mM) conditions. H 2 O 2 dose-dependently increased insulin release under these conditions (Fig. 1A), and this correlated with increases in cellular ROS levels (at H 2 O 2 levels as low as 5 M) (Fig. 1B). Hence, it would appear that, at least in Min6 cells, ROS alone may not be sufficient to stimulate insulin release in the absence of glucose but is required to amplify insulin release under glucose-replete conditions.
Induction of Glutathionylation Amplifies GSIS-It is clear that redox circuits, through ROS signaling, modulate GSIS (26,27). Conjugation of GSH to exposed thiol residues, a covalent modification referred to as glutathionylation, also plays a part in redox signaling (28). Indeed, changes in the glutathionylated proteome coincide with alterations in ROS levels that control many mitochondrial and cellular functions, including aerobic metabolism, signaling, and cell division (29,30). We used diamide, a commonly employed glutathionylation catalyst, to investigate the effect of glutathionylation on insulinoma cell physiology and GSIS (6,31). First, we tested the toxicity of diamide in Min6 cells energized for 1 h with 25 mM glucose because diamide can induce mitochondrial permeability transition and cell death (32). A concentration of 1000 M diamide was required to induce significant increases in ROS and decreases in cell redox potential (Fig. 2, A and B). PI fluorescence, an index of cell death, increased upon exposure of Min6 cells to Ն500 M diamide (Fig. 2C). Acute treatment of Min6 cells with 10 M did not increase GSIS (Fig. 2D). However, exposure of Min6 cells to 100 and 200 M diamide amplified GSIS (Fig. 2D). These concentrations did not induce cell death (Fig. 2, A-C). In glucose-starved cells, 100 M diamide did not stimulate insulin release. However, a significant increase in insulin release was observed when glucose-starved cells were exposed to 200 M diamide (Fig. 2D). The lower concentration of diamide required at high glucose concentrations to amplify GSIS suggests that in glucose-replete conditions UCP2 is in a more deglutathionylated state. We confirmed the GSIS-amplifying effect of glutathionylation using BioGEE, a cell-permeable glutathione molecule tagged to a biotin group. Exposure of glucose-replete Min6 cells to 1 mM BioGEE amplified GSIS (Fig.  2E).
The glutathionylation status of UCP2 and UCP3 dictates the degree of proton leak through either protein, which in turn modulates ROS emission from mitochondria (22). Because mitochondrial bioenergetics play a central role in GSIS and UCP2 is known to negatively regulate this process, we tested the effect of diamide and H 2 O 2 on proton leak-dependent respiration in intact Min6 cells transduced with either scrambled shRNA (control; shCtl) or shRNA directed against UCP2 (UCP2 knock down; shUCP2). Transduction of Min6 cells with shUCP2 lentiviral particles decreased UCP2 protein levels by ϳ81% (Fig. 3A). Cells were incubated in diamide (100 M) or H 2 O 2 (10 M) for 5 min, and then resting, glucose-stimulated state 4 (proton leak-dependent) and extramitochondrial respirations were tested. A summary of the Seahorse XF24 trace generated during the experiment is provided in Fig. 3B. Diamide inhibited and H 2 O 2 activated UCP2 proton leak. In the shUCP2 Min6 cells, the contribution of proton leak-dependent respiration was reduced by ϳ30%. This observation is consistent with previous reports that have shown that UCP2 makes a significant contribution to total respiration in insulinoma cells (20). As shown in Fig. 3C, diamide treatment decreased proton leak-dependent respiration in a UCP2-dependent fashion, suggesting glutathionylation inhibits proton leak through UCP2. However, acute treatment with H 2 O 2 (10 M) had the opposite effect. H 2 O 2 treatment increased proton leak in cells transduced with control shRNA, suggesting that in the presence of small amounts of H 2 O 2 respiration is less coupled (Fig. 3D). We also used TMRE to measure the effect of diamide on mitochondrial membrane potential. TMRE measurements revealed that diamide treatment increased mitochondrial membrane potential in Min6 cells in a UCP2-dependent fashion (supplemental Fig. 1). Conversely, H 2 O 2 (10 M) decreased mitochondrial membrane potential in a UCP2-dependent fashion. Overall, these results indicate that glutathionylation and ROS work in tandem to deactivate and activate UCP2-mediated uncoupling, respectively, in Min6 cells (Fig. 3E).

Assessment of GSIS in UCP2 Knockdown Cells Exposed to
Diamide or H 2 O 2 -The evidence provided above is contradictory because 10 M H 2 O 2 amplified GSIS but also activated leak through UCP2. We next tested the impact of UCP2 knockdown on GSIS and UCP2-dependent ROS-mediated amplification of GSIS. Loss of UCP2 led to a small but significant increase in ATP/ADP, consistent with the notion that UCP2 negatively regulates GSIS by changing mitochondrial coupling efficiency (Fig. 4A). Diamide treatment amplified GSIS in Min6 cells transduced with control shRNA (Fig. 4B). This increase was blunted in cells knocked down for UCP2. In comparison, H 2 O 2 treatment (10 M) increased GSIS in both the shCtl and shUCP2 cells indicating that the amplification of GSIS with H 2 O 2 is independent of UCP2 (Fig. 4B). In glucose-starved cells, diamide and H 2 O 2 also promoted insulin release but only in the cells transduced with control shRNA (Fig. 4B). The observation that diamide augments GSIS by preventing leak through UCP2 prompted us to also test if diamide could have similar effects in mouse islets. Treatment of islets from RIPCre mice with 10 M diamide led to a significant increase in glucose-stimulated insulin release (Fig. 4C). These effects were blunted in islets from mice with ␤ cell-specific UCP2 deletion (UCP2BKO) exposed to 16.7 mM glucose (Fig. 4C). In fact, treatment of islets from UCP2BKO mice with 10 M diamide led to a significant decrease in glucose-stimulated insulin release. No changes in insulin secretion were observed in RIP-Cre or UCP2BKO islets exposed to 2.8 mM glucose (Fig. 4C). These results reveal that diamide is also able to amplify GSIS from mouse pancreatic islets, and this amplification is UCP2specific. Overall, our results indicate that the pharmacological induction of UCP2 glutathionylation amplifies GSIS in both insulinoma cells and mouse islets.

Impact of Glucose Energization on Glutathione Pools, Total Cellular ROS, and Mitochondrial ROS-The observation that glutathionylation impedes UCP2-mediated proton leak in
Min6 cells prompted us to test the glutathionylation status of UCP2 in cells exposed to low or high glucose conditions for 60 min. UCP2 was less glutathionylated in cells exposed to 25 mM glucose signifying UCP2 activation (Fig. 5A). Treatment with 10 M H 2 O 2 increased UCP2 pulldown from cells treated with 25 mM glucose. There appeared to be an increase in UCP2 enrichment in cells treated with 1 mM glucose and 10 M H 2 O 2 , but the difference was small. These results illustrate that UCP2 is less glutathionylated following exposure to 25 mM glucose, which can be enhanced by a brief H 2 O 2 treatment. We also measured the expression levels of glutaredoxin (Grx) 1 and Grx2. Grx1 is expressed in the cytosol and intermembrane space of mitochondria, whereas Grx2 is found in the matrix. Both enzymes display thiol transferase activity and are, to date, the most well characterized enzymes involved in (de)glutathionylation (21). Small increases in Grx1 and Grx2 protein levels were observed in Min6 cells energized with 25 mM glucose (supplemental Fig. 1). The observation that glucose energization deglutathionylated UCP2 prompted us to measure cellular GSH/GSSG and the total amount of GSH associated with the proteome following a 1-h exposure to either 1 or 25 mM glucose. Incubation of cells in 25 mM glucose increased the GSH/GSSG ratio substantially in comparison with cells incubated in 1 mM glucose (Fig. 5B). It is important to point out that despite this increase, the GSH/GSSG ratio was extremely oxidized (ϳ1 and ϳ3.5 in cells treated with 1 and 25 mM glucose, respectively). This is in contrast to the values generated by Pi et al. (12), who found that the GSH/GSSG was actually quite reduced (ϳ50 -70) in rat INS-1 (832/13) cells. In comparison, mouse islets exposed to 20 mM glucose have a GSH/GSSG of ϳ10 (33). To our knowledge GSH/GSSG in Min6 cells has never been reported. Exposure to 25 mM glucose, conversely, substantially decreased the total amount of GSH associated with the proteome (Fig. 5B). This change in the glutathionylated proteome was confirmed by detecting the amount of BioGEE bound to the proteome using avidin-HRP and immunoblot (Fig. 5C). Immunoblotting revealed several proteins at various molecular weights were modified by BioGEE, including a faint band at ϳ34 kDa (which corresponds to the approximate molecular mass of UCP2) (Fig. 5C). This would indicate that upon glucose exposure, cell proteins become deglutathionylated. The oxidized nature of the GSH pool prompted us to determine whether UCP2 was required to protect Min6 cells from an oxidative challenge. Knockdown of UCP2 did not increase total cell ROS levels in cells exposed to high glucose conditions (Fig.  5D). However, treatment with 10 M H 2 O 2 increased cellular ROS levels only in cells knocked down for UCP2 (Fig. 5D). We next decided to test the following: 1) total cellular ROS levels and 2) mitochondrial ROS levels in cells treated with 1 or 25 mM glucose over a 1-h period. Interestingly, exposure to 25 mM glucose led to a gradual increase in matrix ROS (Fig. 5E, top left  panel). A small but significant increase in mitochondrial matrix ROS (measured with MitoSOX) was observed after a 30-min incubation, which then decreased at 60 min (Fig. 5E). This change in matrix ROS from 30 to 60 min is most likely associated to some extent with the activation of UCP2 proton leak (UCP2 was more deglutathionylated after a 60-min incubation in 25 mM glucose). No changes in mitochondrial matrix ROS were observed in cells treated with 1 mM glucose. We also measured cellular ROS changes using DCFHDA over the 60-min period (note that the MitoSOX and DCFHDA measurements were performed as separate experiments). Intriguingly, exposure to 25 mM glucose over the 60-min period significantly decreased total cellular ROS (Fig. 5E). In contrast, ROS levels did not change in cells exposed to 1 mM glucose.
Matrix ROS Activates UCP2 That Impedes GSIS-We then determined whether matrix ROS, generated using paraquat, were able to activate leaks in a UCP2-dependent manner and whether the activation of UCP2 leaks impeded GSIS. Paraquat is a superoxide-generating bipyridine that accumulates in mitochondria in a membrane potential-dependent manner (34) and can be a useful tool to artificially increase mitochondrial matrix superoxide. We first performed a number of dose response assays to identify the appropriate paraquat concentration that can be used to increase superoxide without disrupting mitochondrial function. We had tried a 1-h pretreatment, but no changes in mitochondrial ROS were observed (data not shown). Therefore, cells were preloaded with paraquat for 18 h prior to each assay. MitoSOX and TMRE measurements (performed separately) revealed that 50 M paraquat induced a sharp increase in matrix O 2 . but did not disrupt mitochondrial membrane potential (Fig. 6A). Higher concentrations of paraquat led to further increases in matrix ROS but also progressively altered TMRE fluorescence (Fig. 6A). We next performed Seahorse XF24 analyses on the impact of paraquat on mitochondrial energetics. As shown in Fig. 6B, paraquat (up to 250 M) did not have any effect on resting or glucose-stimulated respiration. Paraquat had a dose-dependent effect on maximal respiration (stimulated by FCCP). Exposure to 100 and 250 Media were then collected and tested for insulin release. Data were normalized to total cellular protein/well. n ϭ 4, mean Ϯ S.E., one-way ANOVA with Fisher's protected least significant difference post hoc test. * and ** denotes p Յ 0.05 and 0.01, respectively, when compared with the 25 mM glucose control. † denotes p Յ 0.05 when compared with the 1 mM glucose control. C, diamide modulates glucose-stimulated insulin release in a UCP2-dependent manner. Islets from control (RIPCre) and pancreas-specific UCP2 knock-out (UCP2BKO) mice were treated with diamide (0 and 10 M) and then tested for insulin release as described under "Materials and Methods." Insulin release was measured following a 1-h exposure to high glucose (16.7 mM) or low glucose (2.8 mM) conditions. n ϭ 4, mean Ϯ S.E., Student's t test. ** denotes p Յ 0.01 when compared with 0 M diamide control.
M paraquat induced a steady decline in the FCCP response. In fact, 250 M paraquat abolished the FCCP effect entirely. However, 50 M paraquat had no effect on the FCCP response (Fig. 6B). Treatment with 50 M paraquat did however increase respiration associated with proton leak. Hence, 50 M paraquat is the optimal concentration to simulate ROS production in the matrix of Min6 mitochondria without detectable disruption of metabolism. In addition, we confirmed that paraquat was accumulating in mitochondria. Min6 mitochondria were collected, solubilized with 1% maltoside, treated with dithionite, and then tested for the presence of paraquat using a UV-visible scan as described by Cochemé and Murphy (34) (Fig. 6C). A peak between 500 and 650 nm was observed only in digested mitochondria from paraquat-treated cells. Because 50 M paraquat increased the leak in Min6 cells, we decided to test if this increase was UCP2-dependent. As shown in Fig. 6D, paraquat increased proton leak only in cells transduced with scrambled shRNA. In contrast, knockdown of UCP2 abrogated the proton leak-activating effect of paraquat. Because paraquat activated the UCP2 leak, we tested if paraquat was able to impede GSIS. Paraquat treatment led to a small but significant decrease in GSIS when cells were treated with 16.5 mM glucose (Fig. 6E).

DISCUSSION
By dissipating the PMF and decreasing mitochondrial coupling efficiency and ROS emission, UCP2 can diminish GSIS. These observations have made UCP2 a potential target for treatment of type 2 diabetes mellitus (1,7). However, the mechanisms governing UCP2 function have remained elusive. Our group has recently shown that reversible glutathionylation controls leak through UCP2 and UCP3 (6,22). In this study, we show that reversible glutathionylation of UCP2 plays an important signaling role in GSIS. Pharmacological induction of glutathionylation with diamide augmented GSIS from insulinoma cells and islets in a UCP2-dependent manner. Conversely, ROS had the opposite effect, but we found that the site of ROS production can have different effects on GSIS. Indeed, matrix ROS, produced by paraquat, activated leak through UCP2. The ROS-mediated activation of leak through UCP2 is associated with the deglutathionylation of the protein. Indeed, in this study we provided evidence that glucose metabolism, which increases matrix ROS, deglutathionylates UCP2. These effects were amplified by co-treatment with H 2 O 2 . We have previously established that ROS is able to deglutathionylate UCP3 (6). ROS-mediated UCP3 deglutathionylation does not proceed spontaneously and requires the presence of a cellular environment. Concentrations of H 2 O 2 in the millimolar range cannot deglutathionylate UCP3 in vitro, but H 2 O 2 in micromolar amounts deglutathionylates UCP3 in intact cells (6). Although we still have not identified the enzyme that mediates this process, it is clear that ROS deglutathionylates UCP2 (and UCP3) with the aid of an as yet to be identified enzyme. Activation of leak subsequently decreased the GSIS signal. Intriguingly, extracellular ROS amplified GSIS indicating different sources of ROS have different cellular effects, e.g. matrix ROS blunts GSIS, whereas cytosolic/extracellular ROS amplifies GSIS. . Impact of PQ on UCP2-dependent proton leak is summarized to the right of the bioenergetic data. All data were normalized to total protein per well. n ϭ 4, mean Ϯ S.D. Student's t test. E, PQ impedes GSIS. Min6 cells were starved and then treated with different amounts of glucose (1-20 mM) for 1 h. Insulin levels in the incubation medium were normalized to total protein amounts/well. n ϭ 4, mean Ϯ S.E. Student's t test.
Hence, insulin-secreting cells most likely rely on a number of redox circuits to modulate GSIS.
Reversible glutathionylation is emerging as an important post-translational modification required to modulate protein function in response to changes in redox status (30). Regulation of protein function by reversible glutathionylation is especially relevant to mitochondrial energetics because mitochondria are a significant source of ROS and have an environment that promotes reversible glutathionylation reactions (35). Despite the importance of mitochondrial energetics in insulin release from pancreatic ␤ cells, the role of glutathionylation in modulating insulin release has never been tested. In this study, we observed that multiple cellular proteins can be modified by glutathionylation, and the extent of this modification depends on the amount of extracellular glucose. To this end, UCP2 was less glutathionylated when Min6 cells were exposed to high glucose. This was amplified by co-incubation in H 2 O 2 . UCP2 is known to partake in the negative regulation of GSIS. However, the observation that the glutathionylation state of other proteins changes in response to glucose metabolism would suggest that redox environment plays an important role in modulating insulin release in general. Redox biology was recently suggested to play a key role in insulin signaling and release (36). Incubation in high glucose substantially decreased the glutathionylated proteome of cells, which was matched by an increase in GSH/ GSSG. The increase in GSH/GSSG is most likely due to the liberation of GSH from surrounding proteins and increased NADPH production from the hexose monophosphate pathway. Despite this increase in GSH/GSSG, the glutathione pool was highly oxidized especially when cells were starved. ␤ cells have relatively low anti-oxidative enzyme expression and therefore may exhaust their GSH pools quite rapidly (37). In fact, insulinoma cells need to be cultured in the presence of powerful reducing agents, like ␤-mercaptoethanol, to maintain a reduced GSH pool and by extension their ␤ cell-like properties (38 -40). In this study, UCP2 became deglutathionylated following a 60-min glucose treatment. Using intact primary thymocytes, we previously showed that pharmacological induction of glutathionylation with diamide deactivated proton leak through UCP2 (6). Similar observations were made herein with Min6 cells. This effect was dependent on the presence of UCP2 (e.g. the leak was not responsive to diamide in UCP2 knockdown cells). Conversely, the deactivation of leak with diamide enhanced the GSIS signal, an effect that was UCP2-dependent. This result indicates that control of leak through UCP2 by reversible glutathionylation is required to control GSIS.
ROS are well recognized "amplifiers" of GSIS. Pi et al. (12) were the first to show that doses of H 2 O 2 as low as 5 M amplified GSIS. Providing ␤ cells with redox cycling molecules like menadione also increases GSIS (13). Other studies have used electron transport chain inhibitors (e.g. antimycin A or rotenone) to artificially enhance cell ROS levels and stimulate GSIS (15). In our opinion, however, use of electron transport chain inhibitors should be avoided when investigating mitochondrial ROS signaling (especially in ␤ cells) because these drugs also compromise ATP production and induce cell death quite rapidly through uncontrolled ROS production and impaired ATP genesis (41)(42)(43). Here, we demonstrate that low doses of H 2 O 2 amplify GSIS but can also induce proton leak through UCP2. The induction of proton leak with ROS is consistent with work published previously by our group, as well as by others (6,16,22,44). The dual effect of ROS on GSIS prompted us to test if matrix-generated ROS had a different effect on GSIS than ROS provided in the extracellular medium. To simulate matrix ROS production, we used paraquat, which generates superoxide by cycling between an oxidized and semi-reduced state (34) and has been used extensively to study oxidative stress in a wide array of biological systems (45,46). After a battery of assays, we established that exposure to 50 M paraquat for 18 h was sufficient to induce an increase in matrix ROS production without appreciably perturbing Min6 cell bioenergetics. We also decided to use paraquat because, in comparison with other drugs that generate ROS by inhibiting the electron transport chain (e.g. rotenone or antimycin A), it takes advantage of complex I activity to generate ROS. At 50 M paraquat we were able to activate proton leak in a UCP2-dependent fashion. Paraquat induced a small but significant inhibitory effect on GSIS, indicating that matrix ROS impedes GSIS and overall that ROS produced in different cellular compartments can have very different effects on insulin secretion.
Glucose metabolism in ␤ cells has been reported to robustly increase ROS levels, which then are thought to amplify GSIS (11). Glucose energization induced a small but significant increase in mitochondrial matrix ROS. In contrast, glucose metabolism led to an immediate suppression of total cell ROS. This latter observation coincides with several other studies reporting that glucose metabolism actually decreases cellular ROS levels (47,48). In fact, evidence generated in other studies attributes this decline in cell ROS to increased glucose flux through the hexose monophosphate shunt pathway that provides a bulk of the NADPH to a cell (47). In our study, we also observed that GSH/GSSG increased following glucose energization, which coincided with the deglutathionylation and activation of UCP2, a key protein involved in controlling cell ROS levels. The decrease in total cell ROS (from 5-60 min of incubation) and the gradual increase in matrix ROS would suggest that two separate pools of ROS in the cell are being modulated following glucose energization. For example, the gradual increase in matrix ROS due to glucose metabolism would trigger the dissipation of PMF by UCP2 proton leak, which is required to desensitize the insulin secretion signal. It is also important to note that an increase in proton leak through UCP2 would also diminish the emission of ROS from mitochondria, which is also a putative amplifier of GSIS (8). Mitochondrial ROS levels were observed to gradually increase during the initial 30 min following glucose provision and then drop back down at the 60-min mark. Hence, the location of ROS genesis determines its impact on GSIS (e.g. extracellular ROS activates although matrix ROS inhibits) and that activation of UCP2 by matrix ROS is used to eventually dissipate the GSIS signal.
UCP2 is expressed in many tissues and has been linked to a number of cell signaling functions (10,49). In previous studies, we provided novel evidence that UCP2 is controlled by glutathionylation (6). Here, we have shown that reversible glutathionylation of UCP2 can play a signaling role in GSIS. Matrix ROS activated UCP2 proton leak, thereby decreasing GSIS, whereas pharmacological glutathionylation of UCP2 enhanced GSIS. These observations are completely consistent with our previous publications (6,22). A glucose-dependent metabolically driven redox circuit is most likely required to regulate insulin release (summarized in Fig. 7). The reducing equivalents derived from glucose oxidation drive mitochondrial synthesis of ATP, which then signals insulin secretion. However, if the provision of reducing equivalents exceeds the oxidative capacity of the mitochondrial respiratory chain, matrix ROS levels increase. We propose that following glucose uptake, UCP2 remains glutathionylated to allow the establishment of a PMF, which drives ATP production and insulin release. Progressive increases in glucose metabolism increase matrix ROS, and the latter induces deglutathionylation of UCP2, thereby activating proton leak and decreasing insulin release. What are the implications of our finding that extracellular ROS enhanced GSIS? Perhaps this is indicative of another source of ROS that is required to activate a completely separate redox circuit dedicated to GSIS amplification. Hence, it is important to consider contributions from the cytosol (e.g. NADPH oxidase) or the intermembrane space of mitochondria. UCP2 activation by matrix ROS could very well prevent ROS emission from mitochondria, thus diminishing the GSIS signal. Indeed, such a scenario would be consistent with previous observations that ROS self-regulate their own production by activating the uncoupling proteins (8,16). Thus, it appears that ROS-mediated redox signaling in ␤ cells is highly ordered and complex.