Mitochondrial clearance of calcium facilitated by MICU2 controls insulin secretion

Objective Transport of Ca2+ into pancreatic β cell mitochondria facilitates nutrient-mediated insulin secretion. However, the underlying mechanism is unclear. Recent establishment of the molecular identity of the mitochondrial Ca2+ uniporter (MCU) and associated proteins allows modification of mitochondrial Ca2+ transport in intact cells. We examined the consequences of deficiency of the accessory protein MICU2 in rat and human insulin-secreting cells and mouse islets. Methods siRNA silencing of Micu2 in the INS-1 832/13 and EndoC-βH1 cell lines was performed; Micu2−/− mice were also studied. Insulin secretion and mechanistic analyses utilizing live confocal imaging to assess mitochondrial function and intracellular Ca2+ dynamics were performed. Results Silencing of Micu2 abrogated GSIS in the INS-1 832/13 and EndoC-βH1 cells. The Micu2−/− mice also displayed attenuated GSIS. Mitochondrial Ca2+ uptake declined in MICU2-deficient INS-1 832/13 and EndoC-βH1 cells in response to high glucose and high K+. MICU2 silencing in INS-1 832/13 cells, presumably through its effects on mitochondrial Ca2+ uptake, perturbed mitochondrial function illustrated by absent mitochondrial membrane hyperpolarization and lowering of the ATP/ADP ratio in response to elevated glucose. Despite the loss of mitochondrial Ca2+ uptake, cytosolic Ca2+ was lower in siMICU2-treated INS-1 832/13 cells in response to high K+. It was hypothesized that Ca2+ accumulated in the submembrane compartment in MICU2-deficient cells, resulting in desensitization of voltage-dependent Ca2+ channels, lowering total cytosolic Ca2+. Upon high K+ stimulation, MICU2-silenced cells showed higher and prolonged increases in submembrane Ca2+ levels. Conclusions MICU2 plays a critical role in β cell mitochondrial Ca2+ uptake. β cell mitochondria sequestered Ca2+ from the submembrane compartment, preventing desensitization of voltage-dependent Ca2+ channels and facilitating GSIS.


INTRODUCTION
While the canonical model of glucose-stimulated insulin secretion (GSIS) from pancreatic b cells with a central role of the mitochondrion has been established for more than 30 years, many details remain unclear. Extensive literature links mitochondrial Ca 2þ uptake to facilitation of insulin secretion [1e6]. However, it is unclear whether the reported effects of elevated matrix-free Ca 2þ on insulin secretion are due to the bioenergetic consequences of the activation of citric acid cycle enzymes [7], the facilitation of plasma membrane oscillations [8], the generation of so-called coupling factors [5,9], or additional currently unrecognized mechanisms. The molecular identification of the components of the mitochondrial Ca 2þ uniporter with the ability to selectively ablate one or more components allows the consequences of disrupted mitochondrial Ca 2þ transport to be investigated in b cells.
The uniporter complex comprises the pore-forming subunit MCU plus up to six regulatory subunits [10]. MCU forms the complex's central pore-forming subunit, while MCU regulatory subunit b (MCUb), essential MCU regulator (EMRE), and mitochondrial uptake 1 and 2 (MICU1 and MICU2) form the accessory components [11]. As part of the Ca 2þ uniporter holocomplex, MICU1 and MICU2 heterodimerize and act to maintain a threshold for mitochondrial Ca 2þ uptake, upholding the mitochondrial Ca 2þ -buffering capacity [12e14]. It has been demonstrated in INS-1 832/13 insulinoma cells that MCU and MICU1 play a role in metabolic coupling; reduced expression of either decreases mitochondrial Ca 2þ uptake and inhibits insulin secretion [8].
Furthermore, b cell-specific Mcu knockout mice display reduced firstphase insulin release in vivo and impaired mitochondrial Ca 2þ uptake, glucose-induced ATP production, and insulin secretion in vitro [15]. However, the function of MICU2 in stimulus-secretion coupling in pancreatic b cells has not been examined. The situation in other cell types is somewhat unclear. In HeLa cells, silencing of Micu2 leads to increased mitochondrial Ca 2þ uptake [16,17], but only at cytosolic levels lower than 5 mM [17]. A similar phenotype is also observed in 1 HEK-293T cells [13,18], while silencing of Micu2 in mouse liver leads to impaired Ca 2þ uptake kinetics [19]. To resolve the role of Ca 2þ uptake, we examined the impact of MICU2 deficiency on insulinsecreting cells and mouse islets as well as in HEK-293T cells.
Both cell lines were cultured at 37 C with 5% CO 2 . Unless otherwise stated, EndoC-bH1 cells were seeded at 2.3 Â 10 5 cells/cm 2 and INS-1 832/13 cells at 1.5 Â 10 5 cells/cm 2 in 24-well plates (Matrigelfibronectin coated or uncoated) and transfected either with siRNA specific to MICU2 (50 nM) or a scrambled siRNA (50 nM). Forty-eight h after knockdown, the growth media were changed to an overnight starvation medium containing 5.6 mM of glucose for INS-1 832/13 cells and 1 mM of glucose for EndoC-bH1 cells. HEK-293T cells were cultured in DMEM containing 1 g/L of D-glucose and supplemented with 10% heat-inactivated FBS, 100 U/mL of penicillin, and 100 mg/mL of streptomycin (PAA) [22]. The cells were plated at 1.5 Â 10 5 cells/ cm 2 in 24-well plates and transfected with either siRNA specific to MICU2 or a scrambled siRNA. All of the assays were performed 48e 72 h after knockdown.

Micu2 knockout mice
We purchased a Micu2 gene trap from the Texas A&M Institute of Genomic Medicine (College Station, TX, USA), bred the mice, and validated the disruption of the Micu2 gene using standard procedures. The Micu2 knockout mice had a B57BL/6J background and were extensively backcrossed. This mouse was previously characterized [23]. A combination of male and female mice was used with no significant differences between the two sexes.  [20]. Insulin secretion was measured with a rat insulin ELISA kit (Mercodia A/B, Sweden) according to the manufacturer's instructions. EndoC-bH1 cells were starved overnight in 1 mM of glucosecontaining growth medium 1 day before the experiment. On the day of the experiment, the growth medium was removed and replaced with 1 mM of glucose-containing HBSS for 2 h followed by 1 h of stimulation with HBSS containing 1 or 20 mM of glucose [24]. Insulin  follows. Five ng of cDNA were mixed with 0.5 mL of gene assay and 5 mL of TaqMan Master Mix and ddH 2 O to a final volume of 10 mL and qPCR was performed using the following program: 50 C for 2 min, 95 C for 10 min, 40 cycles of 95 C for 15 s, and 60 C for 1 min. In all of the runs, a 10x dilution series was produced of one of the samples and used to generate a calibration curve as well as quality control for amplification efficiency. Gene expression was relatively quantified using the calibration curve. The amount of mRNA was calculated relative to the amount of hypoxanthine-guanine phosphoribosyl transferase (HPRT1) as a reference gene.

SDS-PAGE and western blotting
Cells were harvested in RIPA buffer followed by lysis by shaking the samples for 30 min at 950 rpm and 4 C for 30 min. This was followed by centrifugation at 16,000g and 4 C for 20 min to remove cell debris.
Supernatants/lysates were kept at À80 C until use. Then 10e20 mg of protein were separated on mini-PROTEAN TGX pre-stained gels (Bio-Rad) followed by activation of pre-stained Bio-Rad MP imager. The proteins were blotted onto Trans-Blot Turbo Transfer Packs (Bio-Rad), using Bio-Rad semidry blotter, and total protein for loading control was determined from the pre-stains. Membranes were blocked with 10% BSA or 5% skim milk (for mitochondrial complexes IeV experiments) in TRIS-buffered saline (TBS) unless otherwise stated and hybridized with antibodies for MICU2/EFHA1 (Abcam, ab101465), b-tubulin (Abcam, ab6046), a-tubulin (Abcam, ab7291), and mitochondrial complexes IeV (Abcam, ab110413).
2.6. Single-cell live cytosolic ATP and ATP/ADP ratio measurements using perceval and perceval HR Single-cell cytosolic ATP/ADP ratio measurements were carried out in INS-1 832/13 cells using the genetically encoded biosensor Perceval HR [25]. The cells were seeded on poly-D-lysine-coated 8-wellchambered cover glasses (Lab-Tek, Thermo Fisher Scientific) at a density of 70,000 cells/cm 2 . Twenty-four h after seeding, the cells were co-transfected with siRNA and 1 mg of plasmid-encoding Perceval HR (Addgene ID: 49,083) at approximately 50% cell confluency. The cells were further grown for 48e72 h before measurements. On the day of imaging, the cells were pre-incubated at 37 C in 400 mL of buffer P (135 mM of NaCl, 3.6 mM of KCl, 1.5 mM of CaCl 2 , 0.5 mM of MgSO 4 , 0.5 mM of Na 2 HPO 4 , 10 mM of HEPES, and 5 mM of NaHCO 3 at a pH of 7.4) containing 2.8 mM glucose. After 1.5 h of incubation, the cells were imaged at 488 nm excitation and 505e535 nm emission recorded on a Zeiss LSM510 inverted confocal fluorescence microscope. Mouse islets were transduced with adenovirus expressing the similar ATP biosensor Perceval [26,27]. Poly-D-lysine-coated 8-well-chambered cover glasses (Lab-Tek, Thermo Fisher Scientific) containing mouse islets were incubated for 2 h with RPMI with 10% FBS, 100 U of penicillin, and 100 mg/mL of streptomycin containing the Perceval adenovirus. Then this medium was replaced with fresh RPMI medium with 10% FBS, 100 U of penicillin, and 100 mg/mL of streptomycin and incubated overnight. On the day of imaging, the cells were pre-incubated at Original Article 37 C in 400 mL of buffer P (135 mM of NaCl, 3.6 mM of KCl, 1.5 mM of CaCl 2 , 0.5 mM of MgSO 4 , 0.5 mM of Na 2 HPO 4 , 10 mM of HEPES, and 5 mM of NaHCO 3 at a pH of 7.4) containing 2.8 mM of glucose. After 1.5 h of incubation, the cells were imaged as described for clonal cell lines at 488 nm excitation and emission recorded at 505e 530 nm on a Zeiss LSM510 inverted confocal fluorescence microscope using a 40x/0.75 objective. Images were recorded at a frequency of 0.25 Hz (scan time ¼ 3.93 s). Equimolarity was accounted for in the 36 mM KCl solution.
2.7. Single-cell live cytoplasmic-free Ca 2þ measurements Cells were seeded onto poly-D-lysine coated 8-well-chambered cover glasses (Lab-Tek, Naperville, IL, USA) at a density of 70,000 cells/ cm 2 per well. Twenty-four h after seeding, the cells were transfected with siRNA and incubated for further 48e72 h prior to imaging. On the day of imaging, the cells were pre-incubated at 37 C in 400 mL of buffer P containing 2.8 mM of glucose. After 1.5 h of incubation, based on variable Ca 2þ affinities, INS-1 832/13 and HEK-293T cells were loaded with 2 mM of Fluo4 AM and Fluo5F AM, respectively.
Incubation was continued for an additional 30 min. Immediately prior to imaging, the pre-incubation buffer was exchanged with 400 mL of buffer P. Fluo4 AM and Fluo5F AM were excited at 488 nm and emission recorded using a 505 nm emission filter on a Zeiss LSM510 inverted confocal fluorescence microscope using a 40x/0.75 objective. Images were recorded at a frequency of 0.64 Hz (scan time ¼ 1.57 s) using a pinhole diameter of 463 mm. Single-cell-free cytoplasmic Ca 2þ traces were displayed in arbitrary fluorescent units. For experiments performed with the line scan configuration, INS-1 832/13 cells and EndoC-bH1 cells were plated, transfected, and loaded as previously described, except Fluo8 AM (2 mM) used instead in EndoC-bH1 cells. Imaging was again performed using a Zeiss LSM510 inverted confocal fluorescence microscope with a 100x/1.45 objective with Fluo4/Fluo8 excited at 488 nm and emission recorded using a 505 nm emission filter. For each recording, a single line was drawn through the center of 1e3 cells. This was the only part of the field of view that was measured during the recording. The pinhole diameter was reduced to 256 mm to minimize signals from outside the focal plane (1.4 mm section). This allowed the Ca 2þ signals to be tracked as they propagated from the edge to the center of the cells with minimal interference from Ca 2þ diffusing from above or below the focal plane. Recordings were performed at a frequency of 333 Hz (scan time ¼ 3 ms) and a duration of 6000 cycles.

Single-cell live mitochondrial Ca 2þ measurements
Single-cell [Ca 2þ ] mito measurements were carried out using genetically encoded mitochondrial-targeted Ca 2þ biosensor mito-case12 (Evrogen cat# FP992.). INS-1 832/13 and HEK-293T cells were seeded onto poly-D-lysine-coated 8-well-chambered cover glasses (Lab-Tek, Thermo Fisher Scientific) at a cell density of 70,000 cells/cm 2 . Twenty-four h after seeding, the cells were cotransfected with siRNA and 1 mg of plasmid encoding mito-case12 at approximately 50% cell confluency. The cells were further grown for 48e72 h before measurements. On the day of imaging, the cells were pre-incubated at 37 C in 400 mL of buffer P containing 2.8 mM of glucose. After 1.5 h, the cells were imaged using a 488 nm laser and emission recorded at 505e530 nm on a Zeiss LSM510 inverted confocal fluorescence microscope using a 100x/1.45 objective. Images were recorded at a frequency of 0.25 Hz (scan time ¼ 3.93 s).
2.9. Live cell mitochondrial-free Ca 2þ measurements using Rhod2 AM EndoC-bH1 cells were seeded onto poly-D-lysine-coated 8-wellchambered cover glasses at a density of 70,000 cells/cm 2 (Lab-Tek, Naperville, IL, USA). The day after seeding, the cells were transfected with siRNA and incubated for 72 h prior to imaging. On the day of imaging, the cells were pre-incubated at 37 C in 400 mL of buffer P containing 2.8 mM of glucose. After 1.5 h of incubation, 2 mM of Rhod-2 AM (Abcam) was loaded for 30 min at 37 C in buffer P. The cells were then washed once with buffer P to remove excess dye prior to imaging. Excitation of Rhod-2 AM was performed at 543 nm and emission recorded at 590e615 nm on a Zeiss LSM510 inverted confocal fluorescence microscope. For single mouse pancreatic islet mitochondrial-free Ca 2þ measurements, poly-D-lysine-coated 8-wellchambered cover glasses containing single complete or partially digested mouse pancreatic islets pre-incubated at 37 C in 400 mL of buffer P containing 2.8 mM of glucose. After 1.5 h of incubation, 2 mM of Rhod-2 AM (Abcam) was loaded for 30 min at 37 C in buffer P. The cells were then washed once with buffer P to remove excess dye prior to imaging. Excitation of Rhod-2 AM was performed at 543 nm and emission recorded at 565e615 nm. The examination of the effect of glucose on [Ca 2þ ] mito was performed on a Zeiss LSM510 inverted confocal fluorescence microscope using a 100x/1.45 objective. Images were recorded at a frequency of 0.13 Hz (scan time ¼ 7.86 s). [Ca 2þ ] mito in EndoC-bH1, the protocol was the same except that after 1.5 h of incubation, 2 mM of Rhod-2 AM (Abcam) and 2 mM of Fluo8 AM were loaded for 30 min at 37 C in buffer P. Imaging was performed on a Zeiss LSM510 inverted confocal fluorescence microscope using a 40x/0.75 objective with images recorded at a frequency of 0.30 Hz (scan time ¼ 3.36 s).
2.11. Single-cell live mitochondrial membrane potential (Dj m ) measurements INS-1 832/13 cells were seeded onto poly-D-lysine-coated 8-wellchambered cover glasses (Lab-Tek, Naperville, IL, USA) at a density of 70,000 cells/cm 2 . Twenty-four h after seeding, the cells were transfected with siRNA and incubated for 48e72 h prior to imaging. For Dj m measurements, the cells were pre-incubated for 2 h in buffer P with 100 nM of TMRM (Invitrogen) and 2.5 mM of cyclosporin A. After incubation, the cells were washed once with buffer P and then imaging was performed. Under these conditions, Dj m measurements were performed in quench mode [28]. TMRM was present in the medium throughout the experiment. TMRM fluorescence measurements were performed using 543 nm excitation and a 560 nm long pass emission filter on a Zeiss LSM510 inverted confocal fluorescence microscope using a 40x/0.75 objective. Images were recorded at a frequency of 0.64 Hz (scan time ¼ 1.57 s).

Single-cell live Subplasmalemmal Ca 2þ measurements
Single-cell [Ca 2þ ] mem measurements were carried out using PeriCambased genetically encoded inner plasma membrane-targeted Ca 2þ probe mem-case12 (Evrogen cat# FP992.). Specifically, INS-1 832/13 and HEK-293T cells were seeded onto poly-D-lysine-coated 8-wellchambered cover glasses (Lab-Tek, Naperville, IL, USA) at a cell density of 70,000 cells/cm 2 . After 24 h of seeding, the cells were cotransfected with siRNA and 1 mg of plasmid-encoding mem-case12 at approximately 50% cell confluency. The cells were then further grown for 48e72 h before measurements. On the day of imaging, the cells were pre-incubated at 37 C in 400 mL of buffer P containing 2.8 mM of glucose. After 1.5 h, the cells were imaged at 488 nm excitation (8%) and emission recorded at 505e535 nm on a Zeiss LSM510 inverted confocal fluorescence microscope using a 100x/1.45 objective. Images were recorded at a frequency of 0.64 Hz (scan time ¼ 1.57 s).

Mouse islet insulin secretion and insulin content
Islets were picked by hand as previously described [29] under a stereomicroscope and incubated in HBSS containing 2.8 mM of glucose for 2 h. This was followed by 1 h of stimulation with HBSS supplemented with 2.8 or 16.7 mM of glucose. Aliquots of incubation medium were collected and insulin secretion was measured with a mouse insulin ELISA kit (Mercodia A/B, Sweden) according to the manufacturer's instructions. To extract insulin from mouse islets for insulin content measurements, 300 ml of acid ethanol (95% ethanol, HCL (37%), and water) were added to tubes containing 10 islets. The tubes were then sonicated 3 times for 5 s. The samples then underwent 5 cycles of freezing/thawing consisting of 5 min of freezing in ethanol/dry ice followed by 5 min of thawing on ice. The samples were the centrifuged for 5 min at 13.4Âg with 50 ml of supernatant removed into a new tube, stored at À20 C, and then measured using mouse insulin ELISA.
2.14. Glucose and insulin tolerance tests All tolerance tests were performed in anesthetized mice, using 0.25 mg of midazolam (Midazolam Panpharma, Fougères, France) combined with 0.5 mg of fluanisone and 0.02 mg of fentanyl (Hypnorm, VetPharma, Leeds, UK). For IVGTT, mice were starved for 2 h prior to tail vein injection with 1 g/kg of D-glucose. For ITT, unfasted mice were intraperitoneally injected with 0.75 mU/g of human insulin (Novo Nordisk, Clayton, NC, USA). In each case, blood was collected via retro-orbital bleeding at the indicated time points after injection. Upon completion, the samples were centrifuged and plasma collected and stored at À20 C. Insulin and glucose concentrations were then determined by a mouse insulin ELISA kit (Mercodia, Uppsala, Sweden) and Infinity Glucose (Oxidase) Reagent (Thermo Fisher Scientific, Waltham, MA, USA), respectively. Glucose measurements for ITT were determined by an Amplex Red Glucose/Glucose Oxidase Assay kit (Thermo Fisher Scientific, Waltham, MA, USA).

Statistical analyses
All of the experiments were repeated on 3 different days with the exception of ITT, which was performed twice. Data in all of the figures are presented as mean AE SEM, and unless specifically stated; statistical significance was determined by unpaired two-tailed Student's t-tests. For imaging experiments, n in the figures represents the number of cells or regions of interest (ROIs). For islet insulin secretion and IVGT assays, n in the figures represents the number of animals. For all of the statistical comparisons, a p value of less than 0.05 was considered significant.

Functional consequences of MICU2 knockdown in insulinsecreting cells
To understand the role of MICU2 in stimulus-secretion coupling in pancreatic b cells, we silenced MICU2 in INS-1 832/13 cells, a rodentderived insulinoma cell line [20], and in EndoC-bH1 cells, a human embryonic insulin-secreting cell line [21,30]. Our experiments revealed that after 48 or 72 h, small interfering (si) RNA treatment was sufficient to significantly reduce the expression of MICU2 at the mRNA and protein level in both cell lines ( Figure 1AeD). Notably, the expression of MCU and MICU1 was unaffected by MICU2 silencing (Supplementary Figs. 1AeB). Given the proposed function of MICU2 in the uniporter holocomplex, we examined the uptake of Ca 2þ into the mitochondria upon MICU2 knockdown. To this end, we transfected INS-1 832/13 cells with a genetically encoded Ca 2þ biosensor targeted to the mitochondrial matrix, mito-case12 [31]. Our experiments revealed that 16.7 mM of glucose raised [Ca 2þ ] mito and that this increase was attenuated by 48% after silencing of MICU2 (p < 0.0001; Figure 1EeF). Similar findings were observed in EndoC-bH1 cells using another reporter, Rhod2, as these cells are difficult to transfect with plasmids [32]. In these cells, 20 mM of glucose triggered a [Ca 2þ ] mito increase that was attenuated by 57% upon MICU2 silencing (p ¼ 0.04; Figure 1GeH).  Figure 1IeJ). Thus, the diminished mitochondrial response to elevated glucose ( Figure 1EeH) could not simply be ascribed to decreased coupling between glucose metabolism and plasma membrane depolarization. EndoC-bH1 cells behaved similarly to rodent insulinoma cells, in which the maximal [Ca 2þ ] mito elevation following depolarization with 50 mM of KCl was diminished by 90% in MICU2-silenced cells (p < 0.0001; Figure 1KeL). This recording was performed simultaneously with the measurement of [Ca 2þ ] c shown in Supplementary Fig. 3B. Our data suggest that MICU2 plays a role in [Ca 2þ ] mito homeostasis in insulin-secreting cell lines, which would be predicted to impact hormone secretion. Indeed, we found that GSIS was reduced by 41% and 51% at 2.8 (p ¼ 0.0023) and 16.7 mM of glucose (p < 0.0001), respectively, in MICU2-deficient INS-1 832/13 cells ( Figure 1M). In MICU2-deficient EndoC-bH1 cells, basal insulin secretion at 2.8 mM of glucose increased by 46% (p ¼ 0.0245; Figure 1N). However, secretion at 20 mM of glucose did not increase; accordingly, the net fold change (2.8 mMe20 mM of glucose) in insulin secretion was markedly and significantly reduced (1.4-fold vs 3.1-fold; p ¼ 0.0018).

MICU2 deficiency impaired bioenergetic functions in insulinsecreting cells
The mitochondrial membrane potential Dj m is an essential facet of mitochondrial bioenergetics underlying both ATP synthesis and Ca 2þ uptake into the mitochondrial matrix via MCU [33]. The mitochondrial membrane potential is influenced by uptake of Ca 2þ into the matrix [34]. Given the potential regulatory role of MICU2 in this process and the observed attenuation of mitochondrial Ca 2þ uptake, we monitored  . It therefore appears plausible that decreased TCA cycle activation after silencing was the cause of the decreased hyperpolarization. Under basal conditions, unnormalized TMRM fluorescence was stronger in Micu2-silenced cells, indicating that mitochondria were depolarized to a greater extent than in control cells ( Supplementary  Fig. 2C). In light of this, we examined the effect of MICU2 deficiency on glucosestimulated changes in the cytosolic ATP/ADP ratio using live cell imaging of INS-1 832/13 cells expressing the fluorescent ATP/ADP reporter Perceval HR [25]. We found that knockdown of MICU2 significantly reduced the increase in the cytosolic ATP/ADP ratio upon stimulation with 16.7 mM of glucose ( Figure 2C), with the maximal cytosolic ATP/ADP ratio increase diminished by 36% after stimulation with 16.7 mM of glucose in MICU2-silenced INS-1 832/13 cells (p ¼ 0.0013; Figure 2CeD). However, as previously stated, the decreased [Ca 2þ ] mito response to elevated glucose could not be solely ascribed to decreased bioenergetic coupling, since it was also apparent when KCl was elevated. Basal levels of ATP production also appeared to be diminished as evident from reduced unnormalized Perceval HR fluorescence at 2.8 mM of glucose ( Supplementary  Fig. 2D).

MICU2 controlled both mitochondrial and cytosolic Ca 2þ entry
In the next set of experiments, we examined whether MICU2 also influences bulk [Ca 2þ ] c in insulin-secreting cells. Glucose-induced Ca 2þ oscillations were reduced in Micu2-silenced INS-1 832/13 cells ( Supplementary Fig. 3A). However, as this could be the result of lower ATP production rather than a direct effect of MICU2 on [Ca 2þ ] c dynamics, the cells were then depolarized non-metabolically by adding 36 mM of KCl. We predicted that depolarization-evoked [Ca 2þ ] c increases would be more pronounced in MICU2-silenced cells due to reduced buffering by the mitochondria. Surprisingly, we observed that the maximal [Ca 2þ ] c increase upon stimulation with 36 mM of KCl decreased by 48% in MICU2-silenced INS-1 832/13 cells (p < 0.0001;   Figure 1K. Given our findings, we hypothesized that the reduced depolarizationevoked [Ca 2þ ] c increases after MICU2 knockdown (and mitochondrial depolarization) were due to reduced net Ca 2þ entry across the plasma membrane (diminished entry or enhanced extrusion). To functionally elucidate this, we imaged INS-1 832/13 cells using a genetically encoded Ca 2þ biosensor targeted to the inner leaflet of the plasma membrane mem-case12 [35e37]. Figure 3C shows the sensor's plasma membrane localization allowing the subplasmalemmal-free Ca 2þ concentration ([Ca 2þ ] mem ) to be determined. If mitochondrial Ca 2þ sequestration was indeed facilitating net Ca 2þ entry into the cell by removing Ca 2þ from the subplasma membrane compartment, then it would be predicted that cells with defective mitochondrial Ca 2þ transport would show enhanced subplasmalemmal Ca 2þ elevation in response to depolarization by KCl. Indeed, the maximal [Ca 2þ ] mem elevation was 20% greater in MICU2deficient INS-1 832/13 cells than in control cells (p ¼ 0.0118; Figure 3D). The incremental area under the curve (iAUC) was also 43% greater in the MICU2-deficient cells (p ¼ 0.015 vs control; Figure 3E). To establish whether this effect on [Ca 2þ ] mem could be accounted for by restricting MCU activity, we used an alternative strategy to inhibit mitochondrial Ca 2þ transport [38]. In the presence of the ATP synthase inhibitor oligomycin, ATP is maintained by glycolysis and Dj m is retained, or even elevated, allowing mitochondrial Ca 2þ transport to occur normally. Further adding an electron transport inhibitor such as antimycin rapidly collapses Dj m with no a priori effect on (glycolysismaintained) cytosolic ATP production. Mitochondrial Ca 2þ transport, however, being dependent upon Dj m , is abolished [38]. We found that although treating control INS-1 832/13 cells with oligomycin/antimycin did not increase the maximal [Ca 2þ ] mem elevation, the Ca 2þ concentration remained elevated longer at the membrane, with a delayed return to the basal state ( Figure 3F). This was reflected by the iAUC of the effect increasing by 58% compared to oligomycin control (p ¼ 0.0026; Figure 3G). Thus, this treatment had a comparable effect on [Ca 2þ ] mem as MICU2 deficiency. An important proviso is that the capacity of INS-1 832/13 cells to accelerate glycolysis in the presence of oligomycin is limited by their low expression of lactate dehydrogenase [39]; therefore, experiments were performed during the time line before which ATP depletion became significant as indicated by the unperturbed mitochondrial membrane potential during this time (data not shown). Furthermore, using oligomycin alone as a control implied that any effect upon ATP production could not explain the reduced extrusion of [Ca 2þ ] mem in oligomycin/antimycin-treated INS-1 832/13 cells and must therefore have been due to a decrease in mitochondrial Ca 2þ uptake. To further examine whether subplasmalemmal Ca 2þ elevations occur when mitochondrial Ca 2þ uptake is abrogated, we imaged INS-1 832/ 13 cells loaded with Fluo4 using the line scan configuration. This allows repeated imaging across a line through a cell axis as opposed to the entire field of view ( Figure 3H). Crucially, this method allows the higher temporal resolution required for tracking Ca 2þ signal propagation following membrane depolarization. To this end, cell(s) were selected, after which the starvation medium was rapidly replaced with medium containing 36 mM of KCl and the recording started simultaneously. It was expected that MICU2-deficient cells would show enhanced subplasmalemmal Ca 2þ elevation in response to depolarization by KCl, reflected by a stronger fluorescence signal intensity at the edge of the line scan. We found that, upon KCl-induced depolarization, Ca 2þ accumulated at the subplasma membrane compartment in siMICU2-treated INS-1 832/13 cells (Figure 3IeK) with a 1.85-fold higher Ca 2þ signal (measured as iAUC) at the edge of the cell compared to its center (p ¼ 0.0266). Furthermore, there was a 2.5fold higher Ca 2þ response at the submembrane compartment in MICU2-deficient INS-1 832/13 cells compared to controls (p ¼ 0.0006 vs controls; Figure 3K). In contrast, there was no significant difference in the Ca 2þ signal between the edge and center in the control cells (p ¼ 0.8027; Figure 3K). Note that when analyzing the line scan data, the nucleus was avoided when selecting the "center" of the cell. To ensure that the effects on [Ca 2þ ] mem were not the result of offtarget effects of Micu2 siRNA on voltage-gated Ca 2þ channel Comparisons were made with an unpaired two-tailed Student's t-test. In 3L, statistical comparisons were made by a paired two-tailed Student's t-test since variability in short-term incubations was greater. *p < 0.05, **p < 0.01, ***p < 0.001, and ****p < 0.0001. expression, expression of L-type and P/Q-type Ca 2þ channels in Micu2-silenced INS-1 832/13 cells were determined by qPCR: no difference in expression was observed ( Supplementary Fig. 4A). We repeated the line scan experiments in EndoC-bH1 cells but were unable to see any difference between control and Micu2-silenced cells (Supplementary Fig. 4B). Using mem-case12 in Endo C-bH1 cells was not possible due to the cytotoxic effects of the transfection protocols. Collectively, these observations support a model in which MICU2deficient mitochondria in the vicinity of the plasma membrane are less capable of removing Ca 2þ from this compartment. Our results thus indicate not only that MICU2 plays a critical role in controlling MCU and Ca 2þ homeostasis, but also that mitochondria play a critical role in buffering local subplasmalemmal Ca 2þ as previously observed [40], facilitating net Ca 2þ entry into the cell. This may occur either by decreasing cytosolic Ca 2þ -mediated desensitization of voltagedependent Ca 2þ channels in the plasma membrane [41,42] or alternatively by restricting Ca 2þ efflux by plasma membrane Ca 2þ -ATPases [43,44]. The latter was unlikely in the present experimental situation since MICU2 deficiency inhibited an increase in the ATP/ADP ratio, which will limit access to ATP for the ATPases. To assess how this change in Ca 2þ dynamics ultimately affects insulin secretion, INS-1 832/13 cells were exposed to 36 mM of KCl for 15 min: insulin secretion was reduced in MICU2-silenced cells ( Figure 3L).

Micu2-deficient mice showed defective stimulus-secretion coupling and insulin secretion
Our experiments suggested that perturbation of MICU2 led to reduced mitochondrial Ca 2 uptake, resulting in perturbed Ca 2þ entry into the cell and, consequently, defective insulin secretion in both rodentderived INS-1 832/13 and human EndoC-bH1 cell lines. To understand the consequences of MICU2 deficiency in a more physiological setting, we used a global Micu2 knockout mouse [23]. As previously observed in insulin-secreting cell lines, glucose-stimulated (20 mM) maximal [Ca 2þ ] mito elevation was diminished by 57% in Micu2 knockout islets compared to wild-type islets (p < 0.0001; Figure 4Ae B). Note that the KCl response was also diminished ( Figure 4A), similar to that observed in the cell lines. These results suggested that MICU2 most likely plays a similar role in maintenance of [Ca 2þ ] mito homeostasis in primary cells to that observed in cell lines. Furthermore, as [Ca 2þ ] mito homeostasis is critical for cellular bioenergetics, we speculated that the Micu2 knockout might perturb the cytosolic ATP/ADP ratio in primary cells, as was seen in the insulinsecreting cell lines. Therefore, in the next set of experiments, we used an adenovirus expressing the biosensor Perceval, originally described as an ATP/ADP sensor [26], but later shown to report cytoplasmic ATP in mouse islet cells [45]. The maximal Perceval signal was diminished by 50% in Micu2 knockout islets compared with wildtype islets upon stimulation with 20 mM glucose (p < 0.0001; Figure 4CeD). A likely consequence of an abrogated increase in the ATP/ADP ratio in pancreatic b cells is impaired insulin secretion. Batch incubations of Micu2 knockout islets demonstrated that basal (2.8 mM of glucose) insulin secretion was significantly higher compared to that in their wild-type counterparts (p ¼ 0.0427; Figure 4E). However, Micu2 knockout islets failed to release insulin in response to high glucose as efficiently as wild-type islets as illustrated by a 50% reduction in the fold response of insulin secretion (16.7e2.8 mM of glucose; p ¼ 0.0362; Figure 4F). Notably, insulin content was found to be increased in Micu2 knockout islets compared with wild-type islets (p < 0.05; Figure 4G). This effect was most likely due to the accumulation of insulin that was not released since INS1 mRNA expression was similar in wild-type and Micu2 knockout islets ( Supplementary  Fig. 5A). Supplementary Fig. 5B shows the islet secretion data normalized to the insulin content derived from the analysis shown in Figure 4G: in vitro results shown this way reflect the in vivo observations, in that stimulated insulin secretion was abrogated while basal secretion was unchanged. To ensure that any effects were not due to changes in the expression of the mitochondrial complexes, we quantified the protein expression of complexes IeV by Western blotting. There were no significant differences between wild-type and Micu2 knockout islets ( Supplementary  Figs. 5CeD). Our experiments thus suggested that stimulussecretion coupling in b cells in Micu2 knockout mice was defective and underlied perturbed GSIS, reminiscent of that found in EndoC-bH1 cells ( Figure 1N). In the next set of experiments, we examined the consequences of Micu2 deficiency on whole-body glucose homeostasis. Glucose elimination was not significantly different between the Micu2 knockout and wild-type mice ( Figure 4H). However, an intravenous glucose tolerance test (IVGTT) revealed a marked reduction in the initial release of insulin (p ¼ 0.0003; Figure 4I): we observed a 74% and 70% reduction in the plasma insulin levels at 1 and 5 min after glucose administration, respectively, in Micu2 knockout mice compared to wild-type mice. Because insulin secretion was impaired while glucose elimination was unaltered in Micu2 deficiency, we performed an insulin tolerance test (ITT). Both strains of mice responded robustly to insulin, with decreased plasma glucose levels ( Figure 4J). These experiments thus demonstrated that the initial release of insulin in Micu2 knockout mice was reduced, likely caused by a stimulus-secretion coupling defect in pancreatic b cells. Hence, Micu2 knockout mice exhibited deficient GSIS both in vitro and in vivo.
3.5. MICU2 regulation of Ca 2þ in mitochondria, cytosol, and subplasmalemmal space was conserved between cell types Previous studies in neuronal cells showed that inhibition of ATP generation and dissipation of mitochondrial membrane potential led to reduced [Ca 2þ ] c upon stimulation with glutamate or 50 mM of KCl [46]. Therefore, having shown the role of MICU2 in regulating [Ca 2þ ] mito and [Ca 2þ ] c in insulin-secreting cells, we asked whether this mechanism may also be observed in other cell types. High-throughput gene expression assays previously demonstrated that the human embryonic kidney-derived HEK-293T cell line robustly expresses levels of MICU2 mRNA [47]. It has been shown that this cell line also expresses voltage-dependent Ca 2þ channels and exhibits endogenous Ca 2þ currents [48,49]. We therefore used this cell line to address the role of MICU2 in cellular Ca 2þ homeostasis. MICU2 was effectively silenced on the mRNA and protein levels in HEK-293T cells (Supplementary Figs. 6AeC). Measurements of [Ca 2þ ] mito with mito-case12 demonstrated that stimulation with 36 mM of KCl increased [Ca 2þ ] mito in HEK293T cells and that this response was attenuated by 46% after silencing of MICU2 (p < 0.0001; Figure 5Ae B). In addition, time to maximal [Ca 2þ ] mito peak in MICU2-silenced cells was 126 s compared to 107 s in controls after 36 mM of KCl stimulation (p ¼ 0.0087; Supplementary Fig. 6D). [Ca 2þ ] c measurements using Fluo5F showed that membrane depolarization by 50 mM KCl triggered a [Ca 2þ ] c increase and the maximal [Ca 2þ ] c elevation was reduced by 74% in MICU2-silenced HEK-293T cells compared to control cells (p < 0.0001; Figure 5CeD). For experiments measuring [Ca 2þ ] mem using mem-case12, oligomycin was added to ensure that any effects of MICU2 silencing were independent from the resultant bioenergetic effects of increased [Ca 2þ ] mito . The same was not done in the experiments with INS-1 832/13 cells due to ATP production in the b cell being finely tuned to the extracellular glucose concentration as a consequence of the low-affinity enzymes GLUT2 and glucokinase and downstream dissipative pathways (high leak current) [50]. This means that at low glucose, metabolic flux and ATP production in INS-1 832/13 cells are low compared to HEK293T cells. Therefore, any effect of increased [Ca 2þ ] mito on ATP production would be unlikely to reach the threshold to affect K ATP channel activity, meaning using oligomycin was not necessary. Reminiscent of the findings in INS-1 832/13 cells, [Ca 2þ ] mem elevation upon 50 mM of KCl stimulation increased by 35% in MICU2-silenced HEK293T cells compared to the control cells (p ¼ 0.0103; Figure 5EeF). Thus, these data demonstrate a conservation of the regulation of [Ca 2þ ] mito and [Ca 2þ ] c by MICU2 in different cell types.

DISCUSSION
It is undisputed that an increase in [Ca 2þ ] c triggers insulin secretion as well as secretion of most other hormones, but it has also been recognized that such a rise in itself does not sustain insulin secretion [51,52]. Cytosolic Ca 2þ is further transported into the mitochondrial matrix, and an elevation in free matrix Ca 2þ has been considered to trigger additional regulatory factors [53]. Prior to the molecular identification of the components of the uniporter holocomplex [19,54,55], indirect approaches were required to investigate the effect of modulating [Ca 2þ ] mito , none of which were entirely satisfactory. Removal of external Ca 2þ [56] affects multiple non-mitochondrial processes, not least exocytosis itself. Moreover, attempts to dampen changes in [Ca 2þ ] mito by increasing the matrix Ca 2þ -buffering capacity by targeting exogenous Ca 2þ -binding proteins to the matrix [5] or loading cells with Ca 2þ -chelator 1,2-bis(o-aminophenoxy)ethane-N,N,N 0 ,N 0tetraacetic acid (BAPTA) [57] each succeeded in reducing the bioenergetic response to elevated glucose, but concerns remain about toxicity associated with this treatment [50]. The molecular characterization of the uniporter [19,54,55,58] as well as a number of accessory proteins such as MICU1, 2, 3, and EMRE [19,55,59] opened the possibility of genetic manipulation of the uniporter's components [60]. It was shown that silencing of Mcu abrogates the increase in [Ca 2þ ] mito and the cytosolic ATP/ADP ratio in response to glucose in mouse b cells [61]. Another study demonstrated that silencing of MCU in clonal INS-1-E cells resulted in reduced nutrient-induced hyperpolarization of the inner mitochondrial membrane, ATP production, and GSIS [9]. Moreover, knockdown of MICU1 leads to reductions in [Ca 2þ ] mito , cytosolic ATP levels, and GSIS in clonal insulin-secreting cells [8]. However, the role of MICU2 remains more enigmatic. MICU2 is a peripheral membrane protein associated with the mitochondrial inner membrane and facing the intermembrane space. The absence of MICU2 has been associated with reduced Ca 2þ uptake in vivo in mouse liver, though silencing in this system also leads to a decrease in the abundance of the pore-forming subunit MCU and the complex [19]. Observations in other cell systems suggested that the accessory protein may have a regulatory role in Ca 2þ uptake into mitochondria [13,16]. MICU1 and MICU2 are thought to heterodimerize via disulfide bridges and thus regulate Ca 2þ entry into the mitochondrial matrix via the MCU [62]. Indeed, both MICU1 and MICU2 have been shown to play an important role in setting the threshold for mitochondrial Ca 2þ uptake in HeLa cells at variable Ca 2þ concentrations [16]. Furthermore, loss of MICU1 or MICU2 in permeabilized HEK-293T cells has also been shown to disrupt this critical threshold for mitochondrial Ca 2þ uptake [13,14,63,64]. Herein, using MICU2-deficient rodent and human insulin-producing cell lines as well as Micu2 knockout mice, we were able to show that reduced Ca 2þ uptake into mitochondria resulted in inhibition of GSIS. This confirmed observations in previous studies [8,9,61] showing that [Ca 2þ ] mito affects mitochondrial metabolism and insulin secretion. However, our experiments could not resolve how Micu2 knockout mice maintained euglycemia despite reduced early insulin secretion. Insulin tolerance, albeit a rough estimate of insulin action in vivo, was unchanged. In fact, we previously reported that ITT is not sensitive enough to reflect changes in insulin sensitivity in the face of reduced early insulin secretion in vivo [65]. This phenotype is also somewhat similar to what was observed in b cell-specific Mcu knockout mice that exhibit impaired insulin secretion 5 min after intraperitoneal glucose injection with improved glucose tolerance despite normal insulin sensitivity [15]. A hyperinsulinemic euglycemic clamp, which is beyond the scope of the present study, would be required to accurately determine insulin sensitivity. Importantly, MCU and The key observation novel in the context of b cells is that mitochondrial Ca 2þ uptake exerts a profound stimulatory effect on the net flux of cations across the plasma membrane, either by enhancing entry or depressing efflux. Insulin-secreting cells have a high concentration of subplasma membrane mitochondria, with roughly 20% of the membrane inner surface associated with mitochondria exerting a strong Ca 2þ -buffering effect [40]. Mitochondria are dynamic organelles and submembrane Ca 2þ buffering is reduced upon plasma membrane depolarization because of Ca 2þ -dependent remodeling of the cortical F-actin network causing translocation of subplasmalemmal mitochondria to the cell interior [40]. In line with these findings, we demonstrated that restricted mitochondrial Ca 2þ uptake resulted in more pronounced depolarization-induced increases in Ca 2þ in the submembrane compartment. However, under the same conditions, the global cytoplasmic Ca 2þ increases were attenuated. To rationalize a decreased Ca 2þ response to plasma membrane depolarization in both the matrix and cytosolic compartments, the simplest assumption is that net Ca 2þ influx across the plasma membrane is inhibited. Increased extrusion of Ca 2þ is less likely given the observed decrease in the ATP/ADP ratio as a consequence of MICU2 deficiency and that extrusion is an ATP-dependent process. While the endoplasmic reticulum does take up Ca 2þ in response to the elevated ATP/ADP ratio at high glucose, the response is too slow to account for a Ca 2þ deficit [66]. In any case, the effect of restricted mitochondrial Ca 2þ uptake on bulk [Ca 2þ ] c is seen with elevated K þ as well as with high glucose and is thus not directly dependent on bioenergetics. Had this effect only been observed in response to glucose, then the previously postulated positive effect of mitochondrial Ca 2þ uptake on mitochondrial dehydrogenases would have been the most plausible explanation [67]. A paradoxical decrease in [Ca 2þ ] c elevation following inhibited transport of mitochondrial Ca 2þ was previously observed in cultured neurons [38,46]. Since these studies predated the molecular identification of the MCU, mitochondrial Ca 2þ transport was inhibited by mitochondrial depolarization in the presence of the ATP synthase inhibitor oligomycin together with a rotenone or antimycin electron transport inhibitor. The control was oligomycin alone: both control and mitochondrial depolarized cells maintained similar ATP/ADP ratios supported by glycolysis, but the [Ca 2þ ] c transients induced by Ca 2þ entry through NMDA-selective glutamate receptors or voltage-dependent Ca 2þ channels drastically reduced. In the present study, treatment with oligomycin þ antimycin further confirmed a role of mitochondrial Ca 2þ uptake in clearing subplasmalemmal Ca 2þ in insulin-secreting cells, and the observations were similar to those in neurons [38,46]. Importantly, the net neuronal accumulation of 45 Ca 2þ is strongly inhibited by mitochondrial depolarization [38,46], confirming the organelle's ability to control net flux across the plasma membrane. Notably, mitochondrial Ca 2þ uptake has also been implicated in storeoperated Ca 2þ entry, with MCU knockdown in RBL mast cells shown to reduce Ca 2þ influx through Ca 2þ release-activated channels [68]. It was hypothesized that mitochondria are able to deplete a subplasma membrane layer of elevated Ca 2þ [38,46]. Voltage-dependent Ca 2þ channels become desensitized when local Ca 2þ accumulations are not promptly removed [44,69,70]. In addition, it has been shown in chromaffin cells, neurons, and cardiac muscle that mitochondrial sequestration of subplasmalemmal Ca 2þ prevents desensitization of Ca 2þ channels and thus might uphold hormone secretion [71e73]. Furthermore, mitochondria play an essential role in the maintenance of [Ca 2þ ] c amplitude wave propagation via feedback regulation of plasma membrane channel activity in diverse cellular systems, including rat cortical astrocytes and Xenopus laevis oocytes [74,75]. The response of the subplasmalemmal-targeted Ca 2þ probe mem-case12 and line scan imaging in the present study provided support for the existence of such a layer of elevated Ca 2þ [38,46] during b cell depolarization and its depletion by mitochondrial Ca 2þ uptake. Notably, in the mem-case12 experiments, there was a difference between the effects on [Ca 2þ ] mem between Micu2 silencing and treatment with oligomycin/ antimycin, whereby the former caused an increase in the duration of the subplasmalemmal Ca 2þ increase while the latter resulted in an increased magnitude of the response. This may be explained by the different mechanisms by which mitochondrial Ca 2þ uptake is inhibited. When MICU2 is deficient, uptake at low Ca 2þ is likely to increase as the protein plays an important role in suppressing MCU activity under these conditions [11,13,16,17]. Subsequently, when [Ca 2þ ] c is elevated, the mitochondrial buffering capacity decreases. In contrast, oligomycin and antimycin prevent proton transport from the mitochondrial matrix to the mitochondrial intermembrane space, increasing the matrix pH, subsequently collapsing Dj m and inhibiting Ca 2þ entry. These two methods affect the Ca 2þ electrochemical gradient differently and as a result the Ca 2þ uptake kinetics, potentially explaining the differential effects on [Ca 2þ ] mem . Regarding line scan imaging, one could argue that due to its relatively low K d , Fluo4 is unsuitable for measuring [Ca 2þ ] mem , which can reach concentrations of 10e100 mM [76]. This means there is a risk of Fluo4 saturation at the submembrane compartment. This did not seem to occur in the control, where there was little difference in the Fluo4 signal between the edge and center of the cell. In MICU2-silenced cells, however, there was a clear difference between the edge and center of the cell. If saturation was indeed occurring, this would imply that the observed increase in [Ca 2þ ] mem was in fact an underestimation. Moreover, the saturation problems were mitigated against the complementary findings using the lower-affinity mem-case12 probe [36].
The lack of a [Ca 2þ ] mem effect on MICU2-silenced Endo C-bH1 cells may lie in differences in the composition of voltage-gated Ca 2þ channels between the two cell lines. Notably, the L-type Ca 2þ channel current, which is largely responsible for Ca 2þ entry upon high K þ stimulation due to the rapid inactivation of other Ca 2þ channels [77], only accounts for 20e30% of the Ca 2þ current in Endo C-bH1 cells [78] compared to at least 60% of the Ca 2þ current in INS-1 832/13 cells [79]. This means the [Ca 2þ ] mem levels attained in EndoC-bH1 cells in the line scan experiment may still be able to be cleared by MICU2-deficient mitochondria. Thus, any buildup of Ca 2þ in the subplasmalemmal compartment may occur at a slower rate, evading detection in our experiments. This would also explain the smaller effect of MICU2 silencing on depolarization-evoked [Ca 2þ ] c increases in EndoC-bH1. This study shows that mitochondrial Ca 2þ transport has additional consequences for b cells beyond facilitation of the TCA cycle, which may have varying effects on insulin secretion. On the one hand, attenuation of subplasmalemmal Ca 2þ elevations could reduce the triggering of exocytosis in pre-docked vesicles responsible for firstphase insulin release. On the other hand, there is evidence that increased Ca 2þ in the cytosol is required to transport reserve vesicles to the release site, at least in neuronal systems [80]. Based on the KClinduced insulin secretion experiments, the attenuation of mitochondrial Ca 2þ transport leads to a reduction in insulin secretion. First-phase insulin release occurs in the first 5e10 min [81]. Therefore, while there may be an increase in first-phase insulin secretion, an attenuation of the second phase likely caused by decreased transport of reserve vesicles means that insulin secretion decreases over the 15 min time course.
[Ca 2þ ] c increases may also occur as a consequence of endoplasmic reticulum Ca 2þ release. Indeed, there is evidence in b cells of a basal endoplasmic reticulum Ca 2þ leak channeled toward the mitochondria, which stimulates mitochondrial bioenergetics, priming the mitochondria for an increase in glucose [82,83]. This process has been implicated in the generation of cytosolic Ca 2þ oscillations and GSIS [83]. Although ablating Micu2 would also abrogate this process, it appears to be distinct from the mechanism observed herein. The reduction in [Ca 2þ ] mito and the associated increase in [Ca 2þ ] mem were more or less instantaneous upon exposure to KCl and plasma membrane depolarization as opposed to the endoplasmic reticulum Ca 2þ leak, which seems to occur at a constant rate [82,83]. It was previously demonstrated that dynamic changes in the subplasmalemmal mitochondrial population participate in the regulation of local Ca 2þ levels and consequently exocytosis in MIN6 cells [40]. A limitation of our study is that, at this point, we have not obtained sufficient data on whether MICU2 deficiency impacts the localization, morphology, or number of mitochondria in the submembrane compartment. Changes in any of these parameters in MICU2-deficient cells would likely also affect subplasmalemmal and cytosolic Ca 2þ dynamics. Our results may help resolve some of the contradictions in the concept of amplification pathways and K ATP -independent mechanisms of insulin secretion. The search for coupling factors responsible for K ATPindependent insulin secretion began after the discovery that insulin secretion could be provoked by increasing glucose even when the K ATP channel was kept open by diazoxide and the plasma membrane was depolarized by high concentrations of extracellular KCl, thereby opening voltage-dependent Ca 2þ channels [51]. Metabolic coupling factors were postulated to account for the insulinotropic effects of glucose under K ATP -independent conditions and explain why insulin secretion could not be sustained by an elevation of [Ca 2þ ] c alone [52]. However, at least the latter circumstance can be accounted for by the consequences of mitochondrial Ca 2þ uptake that we observed herein: if Ca 2þ is not removed from the subplasmalemmal space by mitochondria, voltage-dependent Ca 2þ channels will be desensitized. A number of metabolites and metabolic pathways have been convincingly implicated in amplification of GSIS [84]. However, most share a similar feature: it has not been resolved how these factors stimulate exocytosis. Moreover, a number of concerns have also been raised about the bioenergetics of the formation of metabolic coupling factors [50].

CONCLUSIONS
In summary, we uncovered a crucial role of mitochondria in maintaining b cell excitability involving sequestering of Ca 2þ into mitochondria located at the vicinity of the voltage-dependent Ca 2þ channels in the plasma membrane that coupled with the well-established bioenergetic effects accounts for the regulatory effect of mitochondrial Ca 2þ uptake on insulin secretion. The recent molecular characterization of the mitochondrial Ca 2þ uniporter holocomplex holds promise for further elucidating the complexities of b cell stimulus-secretion coupling in which the mitochondria are at center stage.