Mitochondrial Proton Leak Regulated by Cyclophilin D Elevates Insulin Secretion in Islets at Nonstimulatory Glucose Levels

Type 2 diabetes ultimately manifests when islets fail to continue hypersecreting insulin to counteract insulin resistance. On the basis of this premise, it is commonly believed that hyperinsulinemia during the fasted state occurs solely as a compensatory response to fasting hyperglycemia induced by insulin resistance. However, high levels of insulin contribute more to insulin resistance than hyperglycemia (1), elevated fasting insulin precedes the development of severe fasting hyperglycemia (2–4), and fasting hyperinsulinemia is one of the most powerful predictors of type 2 diabetes development (5,6). Remarkably, isolated islets from insulin-resistant mice continue to hypersecrete insulin when cultured under very low glucose concentrations (7,8), but the underlying mechanisms are poorly understood. Moreover, mice with reduced capacity to produce insulin (9) or with reduced lipolysis in β-cells (10) are protected from obesity and insulin resistance induced by high-fat diet (HFD) feeding. Likewise, fasting hyperinsulinemia in patients with prediabetes may also be linked to enhanced use of nonesterified fatty acids (NEFAs) by β-cell mitochondria (11,12). These findings support the notion that alterations in β-cell function and insulin content induced by excess NEFAs can drive fasting hyperinsulinemia independently of the presence of fasting hyperglycemia. However, the exact molecular mechanisms by which excess fat can stimulate insulin secretion at nonstimulatory concentrations of glucose (non–glucose-stimulated insulin secretion [NGSIS]) are largely unknown.

Mitochondria are essential for pancreatic β-cell function, integrating fuel utilization with the generation of multiple intracellular signals stimulating insulin release (13). At low glucose levels, mitochondria in pancreatic β-cells preferentially oxidize NEFAs to sustain energy supply and maintain a relatively low release of insulin (11,14). The subsequent rise in blood glucose after a meal catalyzes a shift in fuel preference in β-cell mitochondria to glucose-derived pyruvate as a primary oxidative substrate. Glucose metabolism drives insulin secretion through multiple mechanisms, including generation of the signaling molecules ATP, NADPH, malonyl-CoA, long-chain acyl-CoAs, glutamate, and mitochondrial guanosine triphosphate (GTP) (15–17). In addition to marked fuel flexibility, β-cell mitochondria also exhibit a relatively high level of proton leak compared with other cell types (18,19). Mitochondrial proton leak is defined as the fraction of oxygen consumption that does not contribute to ATP synthesis in coupled mitochondria (20). We have proposed that NEFA-induced mitochondrial proton leak in β-cells may license the generation of signals stimulating insulin secretion regardless of β-cell ATP demand (21), but the mechanism of how NEFAs trigger mitochondrial proton leak in the β-cell is not known.

NEFAs are potent activators of the mitochondrial permeability transition pore (mPTP), an inner mitochondrial membrane channel formed by dimers of ATP synthase that, among other functions, facilitate reentry of protons into the mitochondrial matrix (22,23). Opening of the mPTP is regulated by an array of molecules and proteins, but the most well-documented protein regulator of the mPTP is cyclophilin D (CypD), a matrix peptidyl-prolyl cis-trans isomerase that activates the mPTP (22,24–26). Despite a well-defined role for the mPTP in cell death (24), emerging research has emphasized a more physiological role for CypD and the mPTP (27). A number of studies have characterized the metabolic phenotype of whole-body and tissue-specific CypD knockout (KO) mice. We (28) and others (29) have shown that whole-body deletion of CypD improves glucose tolerance on an HFD, and these mice do not develop fasting hyperinsulinemia on an HFD (29). An mPTP-related proton leak has been identified in islets and β-cells (30), but the precise role of the mPTP in β-cell proton leak and whether the mPTP influences insulin secretion in obesity and prediabetes have not been well studied.

In this study, we show that a primary increase in mitochondrial proton leak is sufficient to stimulate NGSIS in mouse and human islets. An excess NEFA environment in vivo and ex vivo increases islet proton leak and NGSIS by activating the mPTP. Furthermore, we demonstrate that mPTP-mediated proton leak stimulates NGSIS, but not glucose-stimulated insulin secretion (GSIS), and does not require mitochondrial ATP synthesis to secrete insulin. Taken together, we have uncovered a novel pathway that selectively drives pancreatic islet NGSIS, which may be therapeutically exploited to reduce fasting hyperinsulinemia and potentially prevent type 2 diabetes.

Collagenase P, d-glucose, l-leucine, l-glutamine, oleate, palmitate, poly-d-lysine, and diazoxide were purchased from Sigma-Aldrich (St. Louis, MO). Carbonyl cyanide-4-(trifluoromethoxy) phenylhydrazone (FCCP) was purchased from Sigma-Aldrich and Enzo Life Sciences (Farmingdale, NY). Antimycin A was purchased from Enzo Life Sciences. DMSO was purchased from VWR (Radnor, PA). Fatty acid–free BSA was purchased from EMD Millipore (Billerica, MA). FBS was obtained from Life Technologies (Carlsbad, CA). Oligomycin (Oligo) A and BAPTA-AM were obtained from Calbiochem (San Diego, CA). Sodium pyruvate and Matrigel Matrix GFR Phenol Red-free were purchased from Corning Life Sciences (Corning, NY). Seahorse XF96 Spheroid Microplates, Seahorse XF96 Cell Culture Plates, Seahorse XF96 FluxPaks, Seahorse XF Calibrant Solution, and Seahorse XF Base Medium Minimal DMEM were acquired from Agilent Technologies (Santa Clara, CA). BAM15 was a gift from Dr. Kyle Hoehn (University of New South Wales, Sydney, New South Wales, Australia). NIM811 was acquired from Janssen Pharmaceuticals (Horsham, PA).

C57BL/6J mice were obtained from The Jackson Laboratory. Nine-week-old male C57BL/6J mice were put on a custom diet from Research Diets (D09100310) containing 40% fat, 20% fructose, and 2% cholesterol or a control diet (PicoLab Rodent Diet 20 5053) for 4–15 weeks. Eighteen- to 20-week-old male C57BL/6N mice fed an HFD (60% HFD, D12492; Research Diets, New Brunswick, NJ) and control mice fed a chow diet (NIH-31M diet) were purchased from Taconic Biosciences. Whole-body CypD KO and littermate wild-type (WT) mice were produced from heterozygous breeding pairs derived from a male KO mouse (stock #009071, B6;129-Ppif^tm1Jmol/J; The Jackson Laboratory [25]) and a female WT control mouse (stock #101045, B6129SF2/J; The Jackson Laboratory). All animal care was in accordance with National Institutes of Health guidelines and University of California, Los Angeles (UCLA), institutional animal care and use committee approval (UCLA protocol #16–018). Mice were housed at the UCLA Animal Resources Facility in a 12-h/12-h light-dark cycle and maintained at 20–22.2°C. Animals were provided water and food ad libitum until the day of islet isolation.

Islets were isolated from 10–30-week-old mice through collagenase P injection into the bile duct, as previously described (18,31), with minor modifications. Briefly, excised pancreata were digested in a 37°C water bath for 8–10 min, and the digested tissue was washed with RPMI medium containing 1% fatty acid–free BSA. Isolated islets were then cultured overnight at 37°C in 5% CO₂ in mouse islet medium (11 mmol/L glucose RPMI medium + 10% FBS + 100 units/mL penicillin, and 100 μg/mL streptomycin) before experiments.

Human islets were obtained from the University of Minnesota Schulze Diabetes Institute, as previously described (32). Briefly, islets were isolated from living donors with pancreatitis undergoing total pancreatectomy with islet autotransplantation (TPIAT) (see Table 1 for information regarding donors). Respirometry and insulin secretion were conducted at UCLA 1–2 days post-TPIAT.

Human samples were collected between October 2017 and October 2019 from participants in a prospective single-center institutional review board–approved protocol on TPIAT at the University of Minnesota. Informed consent was obtained from study participants for general study participation and for pancreatic tissue analyses.

To make oleate/palmitate (OP) solutions, palmitic acid (P0500; Sigma) was first dissolved in DMSO to make a 51.7 mmol/L stock. Next, oleic acid (112–80–1; Sigma) was dissolved directly in the palmitate stock to make a 102 mmol/L oleate/51.7 mmol/L palmitate DMSO stock. This 2:1 OP stock or DMSO was then complexed to FBS at 57°C while vortexing to produce a 1 mmol/L oleate/0.5 mmol/L palmitate stock with a final DMSO concentration of 0.5–1.0%, as previously described (8).

Intact mouse islets were seeded in V-bottom 96-well plates (Thermo Fisher Scientific) at either 6 islets/well or individual islets in 1 μL/well of Matrigel (8.4 mg/mL protein), and insulin secretion was measured while following a previously established protocol (33). Islets were incubated at 37°C for 3 min to solidify the Matrigel, and 100 μL/well of warm mouse islet medium (11 mmol/L glucose) was added. Islets were incubated at 37°C in 5% CO₂ for ∼45–60 min. Islets were then washed two times with 3 mmol/L glucose Krebs buffer with 0.5% BSA (32) and preincubated in fresh 3 mmol/L glucose Krebs for 30 min at 37°C. This was followed by a 45-min incubation in either 3 mmol/L or 16.7 mmol/L glucose Krebs solution for NGSIS or GSIS, respectively. Nutrients and/or compounds were diluted in Krebs buffer as described in the figure legends. NIM811-sensitive NGSIS in islets was calculated by subtracting the NGSIS values in the presence of NIM811 from the average DMSO control value and dividing this value by the DMSO control NGSIS and multiplying by 100.

An HTRF insulin assay kit (Cisbio, Bedford, MA) consisting of a fluorescence resonance energy transfer–based sandwich immunoassay with Europium cryptate and allophycocyanin (XL665) antibodies was used to quantify insulin levels in the media from insulin secretion experiments (32). Insulin concentration of the samples was calculated from an insulin standard curve (0–20 ng/mL or 0–100 ng/mL).

Seahorse XF24 or XF96 extracellular flux analyzers (Seahorse Bioscience, Billerica, MA) were used for intact mouse islet respirometry while following previously described methods (18,32,34) with minor modifications. For XF96 assays, mouse (1–25 per well) or human (1 per well) islets were seeded in 1 μL/well Matrigel in a poly-d-lysine–coated XF96 plate and size matched between conditions. Islets were incubated at 37°C for 3.5 min to solidify the Matrigel before the addition of 100–150 μL/well Seahorse assay media (XF Base Media Minimal DMEM, pH 7.4, supplemented with 3 mmol/L glucose and 0.1% FBS). Islets were incubated at 37°C for ∼1 h before starting the assay. For port injections, indicated nutrients/compounds were injected in port A (final concentrations in figure legends) followed by Oligo (final concentration 4.5 μmol/L) in port B, FCCP diluted in a mixture of 80% sodium pyruvate and 20% of 1:1 l-leucine/l-glutamine (final concentration 1 μmol/L FCCP in 11.4 mmol/L sodium pyruvate and 2.9 mmol/L each of leucine/glutamine) in port C, and antimycin A (final concentration 2.5 μmol/L) in port D. NIM811-sensitive proton leak in islets was calculated by subtracting the leak in the presence of NIM811 from the average DMSO control value and dividing this value by the DMSO control leak and multiplying by 100. Islets were imaged after Seahorse experiments using an Operetta high-throughput imaging device (PerkinElmer) in brightfield mode with 2× or 10× objectives.

Human islet Ppif mRNA expression data expressed as log₂ robust multiarray average values and donor BMI values were obtained from the National Center for Biotechnology Information Gene Expression Omnibus database (original sources from Taneera and colleagues [35–37]).

Data are expressed as means ± SEM of at least three independent experiments or animals per group or three technical replicates per condition. P values were calculated by unpaired two-tailed t test, one-way ANOVA with Sidak or Dunnett multiple comparisons test, two-way ANOVA with Tukey multiple comparisons test, or linear regression. Statistical significance was set at P < 0.05.

The data sets generated and/or analyzed during the current study are available from the corresponding author upon reasonable request. Components of the Seahorse XF96 individual islet analysis platform used during the current study are commercially available. No novel applicable resources were generated or analyzed during the current study.

Mitochondrial proton leak can be quantified by determining the fraction of islet mitochondrial respiration that is insensitive to mitochondrial ATP synthesis inhibition by Oligo (38) and can be normalized to basal respiration (Fig. 1A). Previous work has suggested that the inherently high proton leak in pancreatic β-cell mitochondria may amplify GSIS (18,19,21). However, it is not known whether mitochondrial proton leak per se is sufficient to increase insulin secretion at nonstimulatory concentrations of glucose (NGSIS) at 3 mmol/L glucose. To address this, we induced proton leak in isolated mouse islets incubated in 3 mmol/L glucose with the small molecule protonophore BAM15 (Fig. 1A), which selectively shuttles protons across the inner mitochondrial membrane without affecting the plasma membrane proton gradient (39). We found that 1 μmol/L stimulates proton leak without completely removing the control over respiration executed by the mitochondrial ATP synthase, as shown by the response of BAM15-treated islets to 1) the ATP synthase inhibitor Oligo (Fig. 1A) and 2) an additional uncoupler FCCP (Supplementary Fig. 1). Treatment with 1 μmol/L BAM15 doubled mitochondrial proton leak compared with DMSO vehicle (Fig. 1B). BAM15 increased islet NGSIS in a dose-dependent manner, with 1 μmol/L BAM15 inducing an approximately twofold increase in NGSIS (Fig. 1C).

To understand the relative mechanistic importance of mitochondrial proton leak in insulin release, we also investigated the interplay between proton leak and GSIS. Stimulatory glucose (16.7 mmol/L) increased mitochondrial proton leak (Supplementary Fig. 1) and elevated insulin secretion almost 30-fold over NGSIS (Fig. 1D). While BAM15-stimulated proton leak increased NGSIS, it caused the opposite effect on GSIS (Fig. 1E). Confirming this conclusion, blockade of mitochondrial ATP synthesis with Oligo decreased GSIS but had no effect on NGSIS (Fig. 1E). The only regulatory node of secretion at which NGSIS and GSIS converged was the K_ATP channel because its inhibition with diazoxide blocked both GSIS and NGSIS (Fig. 1F).

Higher NGSIS observed in prediabetes is believed to be mediated by an increase in intracellular NEFA actions in the β-cell (12,40) as well as by signaling transduced by the plasma membrane free fatty acid receptor GPR40 (41,42). In this context, the links among increased NEFA, NGSIS, and mitochondrial proton leak are unclear. Therefore, we sought to determine whether conditions mimicking hyperlipidemia and insulin resistance in mice could increase mitochondrial proton leak and NGSIS in isolated mouse islets. Acute treatment of mouse islets with a 2:1 mixture of OP significantly increased proton leak (Fig. 2A–C) and NGSIS (Fig. 2B and C). We next aimed to identify the molecular entity responsible for NEFA-induced islet proton leak and NGSIS. NEFAs can trigger the opening of the mPTP (43), an inner mitochondrial membrane channel that facilitates reentry of protons into the mitochondrial matrix (22). The overall structure and components of the mPTP are still under active investigation, but the most well-documented regulator of the mPTP is the mitochondrial matrix isomerase CypD (25,26,44). Binding of CypD to the mPTP promotes pore opening, while inhibition of CypD markedly delays mPTP opening (25).

To determine whether the mPTP was responsible for islet proton leak, we treated islets with the nonimmunosuppressive cyclosporine-derived CypD inhibitor NIM811 (45). Pharmacological inhibition of CypD with NIM811 significantly reduced both proton leak (Fig. 2D and E) and NGSIS (Fig. 2F and G) induced by NEFAs without altering basal or maximal respiration (Supplementary Fig. 2). Importantly, in the absence of NEFAs, NIM811 did not affect NGSIS (Fig. 2H). Furthermore, NIM811 did not alter GSIS from mouse islets (Fig. 2I).

To confirm the involvement of CypD in proton leak–mediated insulin secretion, we isolated islets from mice lacking CypD (CypD KO) (25) or with 50% reduction in CypD (CypD Het). Deletion of CypD is a validated model for inhibition of PTP opening by increasing the threshold for activators to open the pore (46). Islets from CypD KO and CypD Het mice showed similar proton leak (Fig. 3A) and NGSIS (Fig. 3B) compared with WT islets in the absence of NEFAs. NEFAs were unable to induce proton leak in CypD KO islets to the same degree as in WT islets (Fig. 3C). CypD Het islets had a mild but not significant reduction in NEFA-stimulated proton leak (Fig. 3C). Islets from both CypD KO and CypD Het mice had significantly diminished NGSIS with NEFAs compared with WT controls (Fig. 3D). NIM811 only decreased NGSIS in WT but not in CypD KO or CypD Het islets (Fig. 3D), consistent with NIM811 targeting CypD. CypD deletion did not alter maximal or glucose-stimulated respiration (Supplementary Fig. 3), glucose-stimulated proton leak (Fig. 3E), or GSIS (Fig. 3F).

CypD mRNA is increased in islets from subjects with prediabetes and diabetes who exhibit elevated NGSIS (35,36). Therefore, it is possible that CypD in the islet may contribute to the aberrantly high NGSIS observed in diet-induced obesity. To determine whether CypD contributed to basal hypersecretion of insulin in obesity, we first measured mitochondrial proton leak in islets isolated from lean WT mice and WT mice fed an HFD with higher NGSIS (Supplementary Fig. 4). Islets from diet-induced obese mice had significantly higher mitochondrial proton leak than islets from lean controls (Fig. 4A). Remarkably, proton leak showed a strong correlation with NGSIS in islets from HFD-fed but not from chow-fed lean mice (Fig. 4B). Given the role of CypD in NEFA-induced NGSIS, we rationalized that the higher proton leak and NGSIS in obesity may be prevented through pharmacological targeting of CypD with NIM811. Islets from obese mice showed significantly higher NIM811-sensitive proton leak (Fig. 4C) and NGSIS (Fig. 4D) than lean mice.

We used a Western diet as an additional model of diet-induced hyperglycemia and insulin resistance associated with higher NGSIS to further investigate whether proton leak drives NGSIS in islets from obese mice (Supplementary Fig. 4). Similar to the HFD model, islets isolated from mice fed a Western diet showed significantly higher mitochondrial proton leak than mice fed a control chow diet (Fig. 4E), and this leak positively correlated with islet NGSIS only in islets from obese mice (Fig. 4F). Islets from mice fed a Western diet showed significantly higher NIM811-sensitive proton leak (Fig. 4G) and NGSIS (Fig. 4H) compared with islets from chow-fed mice.

Mitochondrial matrix calcium is a potent signal activating mPTP opening (22,23). Depletion of intracellular calcium with BAPTA-AM inhibited NEFA-stimulated proton leak in islets of WT mice but had no effect in islets from CypD KO mice (Fig. 4I). BAPTA-AM had no effects in islets of chow-fed control mice (Fig. 4J and K) but decreased both proton leak and NGSIS of islets from obese mice fed a Western diet (Fig. 4L and M).

To test whether proton leak mediates NGSIS in isolated human islets, we obtained islets derived from living human donors undergoing TPIAT (32). Donor islets had an average glucose stimulation of about fivefold over basal (Fig. 5A). Similar to mouse islets, BAM15 significantly increased NGSIS in human islets incubated at 3 mmol/L glucose (Fig. 5B). NEFAs stimulated proton leak and NGSIS in human islets (Fig. 5C) to a similar extent as observed in mouse islets. NIM811 blocked NEFA-stimulated NGSIS in human islets (Fig. 5D).

We then measured the contribution of CypD to NEFA-mediated proton leak in human islets incubated in 6.7 mmol/L glucose, matching the elevated fasting blood glucose measured in an islet donor with prediabetes. NEFAs increased both proton leak (Fig. 5E) and NGSIS (Fig. 5F) in a manner partially dependent on CypD at 3 mmol/L and 6.7 mmol/L glucose. As expected, fasting C-peptide, a surrogate marker for insulin secretion in vivo, tended to be higher in TPIAT patients with higher BMI (Fig. 5G). Fasting C-peptide was also positively correlated with fasting LDL cholesterol (LDL-C) (Fig. 5H), consistent with high LDL-C being a common hallmark of the metabolic syndrome. Fasting C-peptide (Fig. 5I), BMI (Fig. 5J), and fasting LDL-C (Fig. 5K) were positively correlated with NEFA-activated proton leak in islets from a small cohort of TPIAT patients. On the other hand, NEFA-activated proton leak showed no correlation with fasting blood glucose (Fig. 5L) or HbA_1c (Fig. 5M). Since elevated fasting C-peptide, high BMI, and increased circulating LDL-C are markers predicting a high risk of developing type 2 diabetes, we mined previously published human islet transcriptomics data (35–37) to investigate whether susceptibility of PTP opening driven by CypD changed in the progression of type 2 diabetes. Ppif expression was significantly higher in islets from donors with type 2 diabetes than from donors without diabetes (Fig. 5N). Ppif expression tended to be higher in obese (BMI >27 kg/m²) donors with diabetes than in obese donors without diabetes (Fig. 5O and P).

Pancreatic islets and β-cells within islets show immense heterogeneity in mitochondrial function (32,47,48), a phenomenon that is an important factor in the progression of diabetes (49). Upon closer examination of mouse islets, proton leak varied considerably among islet wells, especially in obesity (Fig. 6A). Similarly, islets from obese mice showed a greater range in insulin secretion up to 2 ng/mL per islet (Fig. 6B), and roughly 66% of obese islets had greater leak or NGSIS than the average chow-fed control islets (Fig. 6C).

Therefore, we hypothesized that a small portion of islets within a population may be responsible for the increased proton leak and enhanced NGSIS observed in obesity. We have previously developed in collaboration with Agilent Technologies a new islet respirometry method (32) enabling the bioenergetic profiling of large-sized individual mouse and human islets. Here, we have augmented this novel technique by embedding individual human islets in Matrigel in wells of a regular XF96 Seahorse cell culture plate (Fig. 6D and E), enabling the profiling of mitochondrial bioenergetics in smaller individual islets from populations of human islets (Supplementary Fig. 5). Our analyses revealed remarkable heterogeneity in proton leak among individual human islets in the absence (Fig. 6F) and presence (Fig. 6G) of NEFAs. Treatment of human islets with NEFAs increased overall proton leak, but not all islets responded to NEFAs (Fig. 6H).

We then measured NGSIS from individual human islets after incubation with 200 μmol/L oleate/100 μmol/L palmitate, close to concentrations observed in obese humans (50). Treatment of individual human islets with OP increased NGSIS compared with FBS DMSO control (Fig. 6I). Similar to proton leak measurements, OP treatment increased the proportion of human islets with relatively higher insulin secretion (Fig. 6J). However, most islets displayed a marginal response to NEFAs (Fig. 6I and J). Taken together, our findings suggest that elevated circulating NEFAs in obesity can activate the PTP in pancreatic islets, which drives increased insulin release at low glucose levels (Fig. 6K).

Fasting hyperinsulinemia occurs well before diagnosis of overt type 2 diabetes, yet no therapeutic interventions currently exist that specifically target this metabolic defect. Drugs that reduce insulin secretion by targeting the K_ATP channel in β-cells (51,52) or mimicking the insulin secretion-blocking hormone somatostatin (53) have shown promise in lowering body weight and improving glucose homeostasis. However, these drugs do not discriminate between NGSIS and GSIS. This scarcity of therapeutic options can mainly be attributed to a lack of knowledge regarding the mechanisms regulating insulin secretion that occur at nonstimulatory glucose levels (NGSIS). In this study, we have identified a molecular mechanism that contributes to NGSIS in both murine and human islets. We found that NGSIS in pancreatic islets is controlled by proton leak through the mPTP and does not require mitochondrial ATP synthesis to stimulate insulin secretion. Proton leak–dependent NGSIS is higher in islets from obese mice and can be inhibited through pharmacological targeting of the mPTP activator protein CypD with NIM811. To our knowledge, this is the first discovery of a druggable pathway that specifically regulates NGSIS and may provide a novel therapeutic approach to prevent fasting hyperinsulinemia.

Pancreatic β-cells and islets have an inherently high mitochondrial proton leak compared with other cell types (18), but the molecular entity responsible for this leak has been elusive for decades. Uncoupling protein 2 (UCP2) was proposed as the mediator of β-cell proton leak (54), but β-cell–specific deletion of UCP2 had no effect on islet proton leak (55). Exposure of islet β-cells to an obesogenic environment results in excessive NEFA availability and mitochondrial calcium overload (56), which is the perfect environment to activate the mPTP (22). Indeed, we observed that the proton leak gained in obesity or in the presence of mPTP-activating NEFAs is sensitive to inhibition by NIM811 and to genetic deletion of CypD. CypD and the mPTP have previously been studied in β-cells in vitro (57) and in vivo (58) but only in islet cell death. The mPTP is involved in more than just the demise of islet cells (30), as several physiological roles have been recently discovered (59).

Our work highlights a pathophysiological role for the mPTP in β-cells regulating NGSIS in obesity. Whole-body CypD KO mice show improved glucose tolerance (28,29), increased glucose uptake into skeletal muscle, and enhanced liver glycogen storage when fed an HFD (28). However, muscle- or liver-specific deletion of CypD failed to recapitulate the metabolic benefits observed in the whole-body KO model (60). Given that whole-body deletion of CypD alters insulin levels in vivo (29) and ex vivo, it is plausible that the improved insulin sensitivity in the CypD KO mice on HFD is due to a reduction in proton leak–mediated NGSIS. However, measurement of islet proton leak and fasting insulin levels will have to be measured in mice lacking CypD specifically in β-cells to verify this hypothesis.

Despite the importance of mitochondrial function for GSIS, it is interesting that proton leak seems to only regulate NGSIS. NEFAs are known to alter mitochondrial function and elevate NGSIS through GPR40 signaling (41,42), but GPR40 signaling contributes to both NGSIS and GSIS pathways. Stimulatory glucose alone increased proton leak in pancreatic islets, but pharmacological and genetic inhibition of CypD only reduced proton leak and NGSIS in the presence of OP, with no effects on glucose-stimulated proton leak or GSIS. Since multiple entities facilitate proton leakage across the inner mitochondrial membrane (38), including the ADP/ATP carrier (61) and nicotinamide nucleotide transhydrogenase (62), it is possible that the proton leak activated by high glucose levels may be mediated through a different molecular target in the β-cell than the leak generated by NEFAs at low glucose levels.

Exactly how proton leak elicits insulin release is still unknown. Unlike GSIS, NGSIS is unique in that mitochondrial ATP synthesis does not play a major role. In obesity, we speculate that the elevated circulating NEFAs at fasting glucose levels activate the mPTP (43), increasing proton leak in β-cell mitochondria and driving tricarboxylic acid cycle flux to maintain mitochondrial membrane potential. This selective engagement of the mPTP is plausible because opening of the mPTP relies on the available metabolic substrate (63). Increased tricarboxylic acid cycle flux may promote production of a mitochondrial signaling metabolite, such as GTP (17,64). It is still unclear exactly how mitochondrial GTP promotes insulin secretion, but this process may depend on GTP-dependent PEPCK conversion of oxaloacetate to phosphoenolpyruvate (65). Phosphoenolpyruvate can be exported from mitochondria and used by pyruvate kinase to generate pyruvate and ATP near the K_ATP channel on the plasma membrane (66).

Currently, mPTP function is mainly studied using pharmacological and genetic tools that target the mPTP regulatory protein CypD. CypD is not a critical component of the pore but instead is a validated regulator that induces pore opening (22,25,26,46). The canonical mPTP inhibitor, cyclosporine A, simultaneously inhibits CypD and calcineurin, a phosphatase that controls inflammatory signaling and is also important for β-cell function (67). NIM811 efficiently blocks the ability of CypD to activate the pore without changing calcineurin (45) but inhibits all cyclophilins to some degree and can have other off-target effects (68). CypD ablation reduces the probability of mPTP activation by increasing the threshold for calcium-induced opening (22,25). Because both CypD deletion and NIM811 had similar inhibitory effects on NGSIS, the mPTP is most likely the major effector of NGSIS rather than other actions CypD might have on mitochondrial function.

Pancreatic islets exhibit remarkable functional heterogeneity in mitochondrial function (47) and insulin secretory capacity (32,69). By modifying our method for bioenergetic profiling of large-sized individual mouse and human islets (32), we have facilitated the measurement of mitochondrial proton leak of individual human islets within the population of multiple living donors. Individual bioenergetics analyses revealed marked heterogeneity in proton leak among individual human islets and highlighted a subpopulation of islets with higher sensitivity to NEFA-stimulated proton leak and NGSIS. Our observation of heterogeneous proton leak complements previous work showing that mPTP activity can vary widely between tissues (60). Since selective populations of islets are altered in the pancreas of individuals with type 2 diabetes (49), it is plausible that certain islets are more susceptible to NEFA-induced leak and NGSIS than others.

Fasting hypersecretion of insulin can drive obesity (9,70), insulin resistance (1), and the progression of diabetes (2–4). Mild reduction in insulin secretion can reverse obesity and improve glycemia (10,71); however, there are currently no small-molecule drugs that specifically target NGSIS without altering GSIS. In this study, we have identified a potential therapeutic target for lowering insulin levels at basal glucose levels in obesity and prediabetes. Of note, metformin, the most commonly prescribed drug for type 2 diabetes, also inhibits the PTP in β-cells (57) and restores NGSIS in rat islets exposed to excess fatty acids (72). With sufficient pharmacological validation and optimization, blockade of proton leak in the β-cell may provide a promising new avenue by which to halt progression of type 2 diabetes.

Acknowledgments. The authors thank Drs. Paolo Bernardi (University of Padova, Padova, Italy), Jude Deeney and Barbara Corkey (Boston University), Marc Prentki (Montreal Diabetes Research Center, Centre de Recherche du CHUM, Montreal, Quebec, Canada), Lisa Norquay (Janssen Pharmaceuticals, Spring House, PA), and Daniel Dagan (Technion Israel Institute of Technology, Haifa, Israel) for valuable feedback and discussion during the preparation of the manuscript. The authors also thank Dr. Kyle Hoehn (University of Virginia, Charlottesville, VA, and University of New South Wales, Sydney, New South Wales, Australia) for generously supplying BAM15 and Janssen Pharmaceuticals (Horsham, PA) for supplying NIM811. Finally, the authors are grateful to Dr. Yijun Chen (UCLA) for helping to coordinate and design the Western diet obesity mouse model.

Funding. N.A. was funded by a fellowship from Kuwait University. This project was supported by a grant from the Kuwait Foundation for the Advancement of Sciences under project code CB17-63 MM-01 awarded to N.A.; National Institutes of Health (NIH), National Institute for Diabetes and Digestive and Kidney Diseases (NIDDK), funding mechanism R01-DK-110181 awarded to R.G.K.; UCLA Department of Medicine Chair commitment and University of California, San Diego/UCLA Diabetes Research Center grant NIH P30-DK063491 awarded to M.L.; and NIH NIDDK funding mechanisms 5R01-DK-074778-10 and R01-DK-099618 and American Diabetes Association funding mechanism 1-19-IBS-049 awarded to O.S.S.

Duality of Interest. This project was supported by a Sponsored Research Agreement from Janssen Research and Development (UCLA code #2016-1342). E.P.T., M.L., and O.S.S. are cofounders of Enspire Bio, Inc., a drug discovery company currently developing inhibitors of proton leak–mediated insulin secretion for treatment of obesity and prediabetes. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. E.P.T. and N.A. wrote the manuscript. E.P.T., N.A., S.B., J.D.W., E.R., S.S., K.E., J.L., and L.S. performed experiments and analyzed data. E.P.T., N.A., M.L., and O.S.S. edited the manuscript and revised the manuscript for resubmission. E.P.T. and S.B. performed experiments during the revision process. S.S., J.J.W., M.L., and O.S.S. contributed to the project conception. M.A., Z.S., and J.J.W. isolated and shipped islets from human patients with pancreatitis undergoing total pancreatectomy with islet autotransplantation at the University of Minnesota. M.D.B. obtained patient consent for human islet studies and evaluated patients before and after pancreas removal and islet transplantation. R.G.K. aided in experimental planning. M.L. and O.S.S. are the guarantors of this work and, as such, had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Prior Presentation. Parts of this study were presented in poster form at the 78th Scientific Sessions of the American Diabetes Association, Orlando, FL, 22–26 June 2018.