A Defect in Mitochondrial Complex III but Not in Complexes I or IV Causes Early β-Cell Dysfunction and Hyperglycemia in Mice Free

The pancreatic β-cell is a highly metabolic cell type relying on mitochondrial metabolism to aid in glucose-stimulated insulin secretion (GSIS). The oxidative phosphorylation (OxPhos) system is critical for ATP production and GSIS. Although recent work has shown that OxPhos is not the primary ATP-producing pathway that closes the K_ATP channels (1,2), the ATP and other secondary metabolite signaling molecules produced by β-cell mitochondria are important in potentiating insulin secretion (3–6). Moreover, β-cell mitochondrial dysfunction is a known hallmark of type 2 diabetes (T2D) pathogenesis leading to downstream activation of cellular stress pathways (7,8).

Stimulus-secretion coupling occurs when glucose is metabolized to pyruvate via glycolysis in the cytoplasm and pyruvate is transported into the mitochondria. The rise in glucose and subsequent glycolysis increases the ATP-to-ADP ratio at the plasma membrane. This locally closes the K_ATP channels, depolarizes the plasma membrane, and initiates insulin secretion (9). Once initiated, mitochondrial OxPhos and metabolism contribute to the sustained insulin response by further metabolizing pyruvate via the trichloroacetic acid cycle, producing reducing equivalents for the electron transport chain (ETC). The ETC is composed of four multisubunit protein complexes I–IV (CI–IV) that generate a proton gradient across the inner mitochondrial membrane. This allows ATP synthesis by complex V (ATP synthase), which, together with the ETC, forms the OxPhos system (10). ATP exits the mitochondria, buffering the cytoplasmic ATP-to-ADP ratio and sustains K_ATP channel closure on the plasma membrane and insulin granule exocytosis (Fig. 1A) (11,12).

Global defects in mitochondrial protein transcription, translation, and consequentially, OxPhos function cause β-cell stress and defects in GSIS in several mouse models (13–16). Importantly, hyperglycemia alone has been shown to specifically downregulate components of OxPhos complexes in a mouse model (17). Of note, ETC subunits are downregulated in human T2D islets (18,19). The purpose of this study was to determine whether defects in the individual OxPhos CI, CIII, and CIV impact pancreatic β-cell function and glucose metabolism by using β-cell–specific inducible knock-out (KO) mouse models. Our data showed that defects in OxPhos cause glucose intolerance and hyperglycemia in mice, but there was a surprising variability in severity and time of onset depending on which complex was defective, with CIII defects being the most diabetogenic. The data presented herein support the hypothesis that individual OxPhos complexes contribute differently to β-cell function and open new lines of research into the role of mitochondrial function in T2D.

Experiments were conducted according to protocols approved by the University of Miami Institutional Animal Care and Use Committee. Mice were housed in a pathogen-free animal facility with a 12-h light-dark cycle and were fed ad libitum on standard chow. Inducible CI, CIII, and CIV KO mice were generated by crossing homozygous floxed NDUFS3 (CI), UQCRFS1 (CIII), or COX10 (CIV) with a transgenic mouse heterozygous for the tamoxifen-inducible Cre recombinase under the mouse insulin I promoter (MIP) construct MIPCre^ERT (20–22). The genes targeted are nuclear encoded components or assembly factors of the complexes. NDUFS3 encodes Ndufs3, a catalytic subunit of CI. The UQCRFS1 gene encodes RISP, which controls one of the catalytic subunits of CIII. COX10 encodes Cox10 which is required for the maturation and stability of CIV.

Mice were maintained on a C57Bl6/J background. Uninduced mice appeared normal. Male NDUFS3^{flox/flox;Cre+}, RISP^{flox/flox;Cre+}, or COX10^{flox/flox;Cre+} mice (8–10 weeks old) were administered tamoxifen via intraperitoneal (IP) injection (dissolved in corn oil with ethanol; 150 mg/kg) for 5 consecutive days. To control for the complex KO, littermate NDUFS3^{flox/flox;Cre+}, RISP^{flox/flox;Cre+}, or COX10^{flox/flox;Cre+} mice were administered vehicle control (corn oil with ethanol; 150 mg/kg IP) for 5 consecutive days. MIPCre^ERT + mice administered eithertamoxifen or vehicle were used as controls for any effects of the tamoxifen or Cre recombinase. Possession of the MIPCre^ERT allele and tamoxifen alone produced no effect on body weight or glycemia (Supplementary Fig. 1).

Blood glucose and body weight were monitored weekly. Glycemia was measured from the tail vein using Contour Next EZ glucose meter and test strips (Bayer). Diabetes was defined by three consecutive readings of ≥250 mg/dL glucose. At 0 and 3 weeks postinduction, blood samples from the tail vein were collected. Nonfasting plasma insulin levels were measured by ELISA (Mercodia).

Islets were isolated from mice 3 weeks postinduction. Islets were collected, washed in Hanks’ balanced salt solution, and lysed in cold lysis buffer (125 mmol/L Tris, pH 7.0; 2% SDS, 1 mmol/L dithiothreitol) with protease and phosphatase inhibitors (Roche). Then, 5 μg of total protein was run on 4–20% SDS-PAGE gels. Blots were blocked for 1 h at room temperature with 5% nonfat dry milk in 0.1% Tween-20 in PBS (PBST) and incubated with primary antibodies overnight at 4°C. Horseradish peroxidase-conjugated secondary antibodies (1:2,000) were used, and signal was developed with chemiluminescence. Antibodies are summarized in Supplementary Table 1.

Glucose tolerance was assessed 3 weeks postinduction via IP glucose tolerance tests (IP-GTTs). Mice were fasted overnight (16 h). Glucose (20% w/v, 2 mg/kg) was administered via IP injection, and glycemia was measured periodically. At baseline and 15 min postglucose, blood samples were collected, and plasma was analyzed by ELISA to determine GSIS (Mercodia) and C-peptide secretion (Alpco). To determine peripheral insulin sensitivity, mice were fasted for 6 h at 3 weeks postinduction. Mice were injected with 0.75 units/kg insulin (Humulin, Eli Lily), and glycemia was measured periodically after injection.

Islets were isolated from mice at 3 weeks postinduction, as previously described (23). Briefly, Liberase (0.17 mg/mL, Roche) was injected into the common bile duct. Pancreata were excised and digested at 37°C for 15 min, followed by separation by Histopaque (Sigma-Aldrich) with density centrifugation. Islets were handpicked and cultured for up to 2 days in CMRL medium supplemented with 10% FBS, 1× penicillin/streptomycin, HEPES, and GlutaMAX (1×).

Approximately 75 islets per sample were placed into a column, connected to a PERI5 perifusion system, and perifused at 100 μL/min (Biorep Technologies, Miami Lakes, FL). Islets were flushed for 90 min with HEPES-buffered solution containing 3 mmol/L glucose and were perifused with 3 mmol/L glucose (10 min), 16.7 mmol/L glucose (20 min), 3 mmol/L glucose (20 min), 25 mmol/L KCl (5 min), and 3 mmol/L glucose (10 min). Insulin secretion was determined by ELISA (Mercodia).

Isolated islets were transduced with 1 μL of adenoviral cytosolic GCamp6s (SignaGen Laboratories; Ad-CMV-GCamp6s, 1.85E+10 plaque-forming units/mL) or mitochondrial GCamp6s (University of Iowa Viral Vector Core; Ad5-CMVmitoGCamp6T2A, 2E+11 plaque-forming units/mL) for 48 h after isolation. High glucose (17 mmol/L) and KCl (25 mmol/L) were both applied. For [Ca²⁺] imaging, a Z-stack of 7–11 confocal planes was acquired every 5 s using a ×20 water immersion objective (numerical aperture 0.8) on a Leica SP8 confocal laser-scanning microscope (Leica).

At 3 and 5 weeks postinduction, mice were anesthetized with ketamine/xylazine. Dissected pancreata were fixed and immunostained as previously described (24). Antibodies used are summarized in Supplementary Table 1. Immunostaining was visualized by using Alexa Fluor conjugated secondary antibodies (1:500) in PBS. Confocal images (pinhole = Airy 1) of randomly selected islets were acquired on an inverted laser-scanning confocal microscope (Leica, TCS SP5) with a ×63 oil immersion objective and numerical aperture 1.4. Images were analyzed using ImageJ software (National Institutes of Health, https://imagej.nih.gov/ij/).

Samples were fixed in 2% glutaraldehyde in 0.05 mol/L phosphate buffer and 100 mmol/L sucrose, postfixed overnight in 1% osmium tetroxide in 0.1 mol/L phosphate buffer, dehydrated through a series of graded ethanols, and embedded in a mixture of EM-bed/Araldite (Electron Microscopy Sciences). Then, 1-μm-thick sections were stained with Richardson’s stain for observation under a light microscope. A Leica Ultracut-R ultramicrotome was used to cut 100 nmol/L sections, which were stained with uranyl acetate and lead citrate. The grids were viewed at 80 kV in a JEOL JEM-1400 transmission electron microscope (TEM), and images were captured with an AMT BioSprint digital camera.

Isolated islets were kept in culture for 2 days and then mitochondrial respiration was determined via Seahorse Extracellular Flux Assay using the XFp platform (Agilent). Approximately 30 islets per sample were loaded into the XFp plate. Vo₂ in response to glucose (17 mmol/L), oligomycin (5 μmol/L), carbonylcyanide-4-(trifluoromethoxy)-phenylhydrazone (FCCP; 1 μmol/L), and antimycin A/rotenone (2.5 μmol/L/5 μmol/L) was measured as described previously (25). Data were normalized to basal respiration.

For RNA expression, total RNA was extracted from islet samples using the RNeasy isolation kit (Qiagen). Gene expression was performed by quantitative real-time RT-PCR using Power SYBR Green PCR Mix (Applied Biosystems) with the QuantStudio 3 Real-Time PCR system (Applied Biosystems) with a standard protocol, including a melting curve. Relative abundance for each transcript was calculated by a standard curve of cycle thresholds and normalized to 18S. Primers were purchased from IDT Technologies. Primer sequences are available in Supplementary Table 2.

To quantify changes in Ca²⁺ levels ([Ca²⁺]), we manually selected regions of interest around individual islet endocrine cells. We measured changes in mean fluorescence intensity using ImageJ. Changes in fluorescence intensity were expressed as the percentage change over baseline (ΔF/F). The baseline was defined as the mean of the intensity values during the nonstimulatory period (first 3 min of recording). Raw Ca²⁺ data were detrended and converted to % ΔF/F using a custom MATLAB script (MathWorks). Data were displayed as heat maps to include all viable, recorded cells and are separated by animal. β-Cells were distinguished from other endocrine cells based on morphology, location in the pancreatic islet, and response to glucose. We reported percentages of glucose-responsive per islet per animal cells based on analysis of individual Ca²⁺ traces. Responding cells were considered those that had a mean fluorescence intensity >2 times the SD of baseline fluorescence after stimulation with 17 mmol/L glucose. Due to the adenoviral constructs used, labeling of cells was limited to superficial islet cells. Response to KCl was used to exclude any nonviable cells from analysis.

To quantify loss of complex-specific staining in the KO samples, four to eight randomly selected islets per sample were imaged. ImageJ was used to count the number of complex-positive and insulin-positive cells and the total number of insulin-positive cells per islet. The number of complex-positive and insulin-positive cells was divided by the total number of insulin-positive cells per islet to yield a proportion of insulin-positive cells also stained for the complex of interest. To determine the average islet area, four to eight islets per sample were selected, based on being in the equatorial plane, and imaged. The mean islet area per sample was determined from three confocal planes per islet using ImageJ. Whole pancreas sections were immunostained and scanned, from which the insulin-positive area was determined (no differences, data not shown). The number of insulin-positive (β) and glucagon-positive (α) cells was manually counted, and the ratio was determined. To determine superoxide dismutase 2 (SOD2) and peroxide reductase 3 (Prdx3) staining intensity, we analyzed three confocal planes per islet (four to eight islets per mouse). The whole islet was a region of interest, and we used the “mean gray intensity” function in ImageJ. We normalized the value to total islet area.

Statistical comparisons were performed using unpaired Student t test, Mann-Whitney test, or one-way ANOVA, followed by multiple-comparison procedures with the Tukey or Dunnett tests. Statistical analysis of IP glucose and insulin tolerance tests was performed using two-way ANOVA, followed by the Šidák multiple comparison test. Data are shown as mean ± SD or SEM, as specified in each figure legend. GraphPad Prism 9 software was used for all analyses, and P < 0.05 was considered significant. Experiments were not randomized or blinded.

Further information and requests for resources, reagents, and data should be directed to the corresponding authors and will be fulfilled upon reasonable request.

We achieved successful β-cell–specific OxPhos complex KOs by crossing tamoxifen-inducible MIPCre^ERT–positive mice with mice homozygous for the floxed allele of Ndufs3 (CI), Uqcrfs1 (CIII), or Cox10 (CIV). Each of the genes targeted are nuclear encoded subunits or assembly factors that are essential for the assembly and function of the respective complexes (26–28). Quantitative PCR analyses of total islet DNA from all groups shows that the recombination efficiency of the Cre-LoxP system in these models is high and comparable between groups (Supplementary Fig. 2).

Using immunohistochemistry and Western blot analysis, we determined that at the experimental time point of 3 weeks posttamoxifen induction, all three models had a robust and similar level of protein reduction (Figs. 1 and 2). Immunofluorescent staining for each of the target proteins (NDUFS3, UQCRFS1, and Mt-CO1, which depends on COX10 for maturation and stability) showed a loss of complex staining in most cells in the islet core, while cells in the islet mantle still had complex-specific staining (Fig. 1B–E). Costaining with insulin confirmed that the former were β-cells, and image analysis showed a significant decrease in insulin-positive cells that were also positive for the specific OxPhos complex (Fig. 1B and Supplementary Fig. 3).

Importantly, each model showed a significant and comparable level of decrease in the target protein without changes in other OxPhos complexes (Fig. 2A–C). Based on the quantitative PCR, Western blot, and immunohistochemistry analyses, most β-cells were KO for the respective floxed alleles.

We observed that CIII KO mice developed hyperglycemia 3 weeks after tamoxifen induction, with ∼80% of the mouse population being diabetic at this time point (Fig. 3A and Supplementary Fig. 1B). CIII KO mice had significantly elevated nonfasting and fasting glycemia 3 weeks after tamoxifen induction (Fig. 3B and Supplementary Fig. 1B). At this time point, none of the CI or CIV KO mice were diabetic, and only 25% became diabetic at 8 weeks posttamoxifen (Fig. 3A and Supplementary Fig. 1B). CI and CIV KO mice examined at 12 weeks postinduction showed mild increases in nonfasting glycemia (CIV KO mice) or became hyperglycemic (CI KO mice) but not as severely as CIII KO mice (Supplementary Fig. 1). Body weight was unaffected for any group at any time point (Supplementary Fig. 1A).

To investigate glucose homeostasis, we performed IP-GTTs on all models at the 3-week time point. CIII KO mice were glucose intolerant, but no differences in the other groups compared with their respective vehicle controls were observed (Fig. 3C and D). We measured insulin secretion during the IP-GTTs, and CIII KO mice were unable to secrete insulin in response to the glucose stimulation compared with CIII controls (Fig. 3E). We also measured C-peptide during the IP-GTTs and saw a similar effect, with CIII KO mice showing significantly decreased and delayed C-peptide secretion compared with CIII controls (Fig. 3F). Ex vivo insulin secretion on isolated islets showed no differences between any of the groups analyzed (Supplementary Fig. 4). We observed a decrease in nonfasting plasma insulin levels in CIII KO compared with vehicle control mice (P = 0.0143, unpaired two-tailed t test) (Supplementary Fig. 5A). No differences in insulin tolerance or peripheral insulin sensitivity between CIII KO and CIII control mice were observed (Supplementary Fig. 5B).

To gain further insight into the physiological defects induced by the OxPhos complex KOs in β-cells at the 3-week time point, we measured cytoplasmic Ca²⁺ ([Ca²⁺]_i) using a GCaMP6s viral vector. We monitored [Ca²⁺]_i in intact, isolated islets from MIPCre^ERT controls, CI vehicle, and KO and CIII vehicle and KO mice by live confocal imaging and analyzed and plotted changes in fluorescent intensity for individual cells per islet per animal across samples. Islets were continuously perifused in an imaging chamber and were exposed to basal glucose (3 mmol/L), high glucose (17 mmol/L) and KCl (25 mmol/L, positive control). No differences were observed for any parameter analyzed between the MIPCre^ERT vehicle and tamoxifen-treated islets (Fig. 4A). Interestingly, CI and CIII KO islets both show a significant decrease in peak glucose response compared with their respective controls (Fig. 4B). Importantly, CIII KO islets have a significantly shorter latency to glucose peak compared with CIII vehicle controls (Fig. 4B). These data indicate that although there were no differences in the number of responding β-cells, the magnitude and dynamics of the responses varied depending on the complex targeted.

To investigate the mitochondrial phenotype of the OxPhos-deficient mice, we performed Seahorse extracellular oxygen flux analysis on isolated islets from KO and control mice (Fig. 5). Vehicle controls were pooled for analysis. We determined glucose-stimulated respiration and maximum mitochondrial respiration after FCCP uncoupling. All three OxPhos complex KO models had a similar defect in glucose-stimulated Vo₂ (Fig. 5B), while maximum mitochondrial respiration was unaffected (Fig. 5C). MIPCre^ERT tamoxifen-treated mice showed no differences compared with vehicle controls (Supplementary Fig. 6A and B). No changes were observed in proton leak or CV-dependent respiration between any group (Supplementary Fig. 6C and D).

The similar defects in glucose-stimulated Vo₂ contrast with the pronounced differences in the in vivo phenotypes of the three KO models. We therefore investigated mitochondrial function further by measuring mitochondrial free Ca²⁺ ([Ca²⁺]_mito) with a mitochondrially targeted GCamp6s viral vector (Supplementary Fig. 7A and B). This construct also has a pH red biosensor that we used to validate that changes in pH did not affect mitochondrial Ca²⁺ responses (Supplementary Fig. 8). Changes in confocal fluorescent intensity were analyzed and plotted for individual cells per islet per animal. Islets were continuously perifused in an imaging chamber and were exposed to basal glucose (3 mmol/L), high glucose (17 mmol/L), and KCl (25 mmol/L, to control for excitability and viability).

In contrast to the Seahorse data where all models had a similar defect, stark differences in [Ca²⁺]_mito handling were observed between the models. No differences were observed between MIPCre^ERT vehicle- and tamoxifen-treated islets (Supplementary Fig. 9). Islets from CI or CIV KO mice do not show differences compared with their respective controls in their glucose responses. However, CIV KO islets do have a significantly decreased KCl response. Islets from CIII KO mice showed a significant decrease of glucose-responsive cells, and the cells that did respond to glucose showed a dysregulated response profile and loss of oscillatory behavior (Fig. 6). These data indicate that defects in specific OxPhos complexes contribute uniquely to β-cell mitochondrial dysfunction.

Given the robust in vivo phenotype and dysfunctional [Ca²⁺]_mito and [Ca²⁺]_i, we sought to determine whether islet structure and morphology were altered in mice with a defect in CIII. We examined insulin and glucagon immunostaining in the pancreas of all groups (Fig. 6 and Supplementary Fig. 10). At 3 weeks postinduction, CIII KO islets showed normal structure and ratios of β-cells to non–β-cells (Fig. 7C and D). Only at the 5-week time point we did observe decreases in islet area and loss of β-cell mass in CIII KO islets (Fig. 7C and D). However, upon TEM analysis, we observed changes in mitochondrial morphology in the CIII KO β-cells compared with CIII vehicle control cells at the 3-week time point (Fig. 7E and Supplementary Fig. 11). No differences were observed in the islets of the CI and CIV models and the MIPCre^ERT controls (Supplementary Fig. 10). We did not detect changes in cell death or immune cell infiltration in islets of CIII KO mice at the 3-week time point (Supplementary Fig. 12).

Because we did not observe changes in islet morphology, cell death, or local inflammation at 3 weeks posttamoxifen, we investigated stress pathways and compensatory mechanisms that could be contributing to the CIII phenotype. SOD2 and Prdx3 are localized to the mitochondria. Changes in their expression have been associated with alterations in cellular redox status (29,30). Immunostaining for SOD2 and Prdx3 conducted on all models and their controls showed that only CIII KO islets had a significant increase in SOD2 and Prdx3 intensities (Fig. 8A–H and Supplementary Fig. 13). mRNA analysis showed CIII KO islets had significantly increased expression of SOD2 and catalase (Fig. 8I). Additionally, several critical β-cell genes and disallowed genes in the CIII model were analyzed. We did not find differences in Ins1 or MafA mRNA expression; however, we did observe significant increases in both critical and disallowed genes (Pdx1, Gck1, Glut2, LDHA, LDHB, and MCT1) in the CIII KO islets (Fig. 8I). Individual gene analyses and P values are shown in Supplementary Fig. 14.

Although mitochondrial metabolism and oxidative respiration are crucial for β-cell function and GSIS, the contribution of individual OxPhos complexes to β-cell function has not been explored. We used β-cell–specific KO models of CI, CIII, and CIV to address this question. Our data indicate that deletion of individual complexes affects β-cells and glucose homeostasis uniquely. All models had comparable levels of protein KO and mitochondrial respiratory defects at 3 weeks postinduction. Despite these similar respiratory phenotypes, CIII KO mice developed rapid and robust hyperglycemia and diabetes shortly after tamoxifen-induced KO. This contrasts with the other two models (CI and CIV KO), which developed diabetes much later and less robustly. Although CI could be compensated by increased electrons flow through CII, this feature would not explain why the CIV model phenotype was also milder.

In neuron-specific mouse models of OxPhos complex deficiencies, we similarly showed phenotypic differences between the CIII KO and the CIV KO models (20). The neurological phenotypes also varied widely, with the CIII model leading to sudden death and the CIV model inducing slower but progressive neurodegeneration and behavioral changes. Mouse models for different complex deficiencies thus produce a broad spectrum of pathogenic mechanisms with diverse pathological outcomes (31), highlighting the need to define the effects for each individual respiratory chain defect in the cell of interest. Here, we demonstrate that for the β-cell, whose demise is central to any model of diabetes pathogenesis, deficiencies in CIII had the strongest pathogenic impact.

Our results show that early after induction, CIII KO mice did not secrete insulin or C-peptide in response to a glucose challenge and were glucose intolerant. This was coupled with a disruption of Ca²⁺ signaling and increased levels of antioxidant markers in β-cells. β-Cells increase antioxidant pathways as a defense mechanism against increased oxidative stress and subsequent apoptosis (32–34). The higher immunostaining intensities and mRNA expression of antioxidant enzymes in β-cells of the CIII KO model therefore suggest an early response to compensate for increased oxidative stress. Interestingly, no differences were seen when oxidative stress markers and reactive oxygen species production were investigated (not shown), suggesting that the increased antioxidant response of CIII KO islets compensates for a putative oxidative stress. The fact that H₂O₂ is also an important signaling molecule for insulin secretion also raises the alternative hypothesis that the compensatory responses in β-cells may lead to a reduction in H₂O₂, causing defective insulin secretion. Our TEM images also support that β-cells are still able to cope with the putative stress at an early time point. Although we did not observe alterations in gross islet morphology or cellular composition until 5 weeks after induction, TEM imaging revealed that CIII KO β-cells have altered mitochondrial morphology (larger, more circular). We conclude that CIII KO islets undergo functional dysfunction prior to β-cell loss.

Ca²⁺ signaling and flux through the cytoplasm and organelles, such as the mitochondria and endoplasmic reticulum, is crucial for insulin granule exocytosis and sustaining insulin release (35,36). Here we show that cytosolic and mitochondrial Ca²⁺ responses to glucose differ greatly, depending on the OxPhos complex affected. Cytosolic Ca²⁺ peak glucose responses were impaired in the CI and CIII KO model. However, only the CIII KO had a significantly shorter time to peak glucose. These changes in cytosolic Ca²⁺ responses are hallmarks of β-cell dysfunction seen in other mouse models (35,37–40). The early Ca²⁺ response could be a symptom of diminished Ca²⁺ buffering by the endoplasmic reticulum and mitochondria (41–43). Indeed, a significant decrease in glucose-responding cells in mitochondrial Ca²⁺ was observed for the CIII KO. It is likely that this cytosolic Ca²⁺ defect alters the mitochondrial Ca²⁺ response to glucose, affecting insulin secretion in vivo (36,44). We propose that CIII KO mice have this severe Ca²⁺-handling phenotype due to the their altered cellular redox status (45).

A defective mitochondrial Ca²⁺ homeostasis may affect multiple pathways that ultimately impact insulin secretion and β-cell survival (46). Indeed, mitochondrial Ca²⁺ is an important signal for activating mitochondrial energy metabolism, and blunting mitochondrial Ca²⁺ impairs insulin secretion (37,38,44). Recent studies have shown that KO of the pore-forming subunit of the mitochondrial Ca²⁺ uniporter complex in mouse β-cells causes insulin secretion deficits and completely disrupts mitochondrial Ca²⁺ responses to high glucose (35). Of pathological relevance, the mitochondrial Ca²⁺ uniporter complex, together with the mitochondrial membrane potential, tightly regulates the influx of mitochondrial Ca²⁺ to prevent apoptosis (47). Also of note are the significantly increased disallowed β-cell genes and increased glycolytic markers in the CIII KO islets (glucokinase, lactate dehydrogenase A, etc.).

The fact that we observe normal ex vivo insulin secretion in the CIII KO model, despite having a dramatic in vivo insulin secretory defect, is indeed paradoxical. One hypothesis is that CIII KO islets have increased glycolysis to compensate for their decreased mitochondrial function that compensates their ex vivo insulin secretion. We observed increased Glut2 and glucokinase mRNA expression, indicating increased glycolytic flux. Additionally, CIII KO islets have increased lactate dehydrogenases and the lactate transporter. Notably, a recent study showed that lactate dehydrogenase is part of a plasma membrane localized NAD⁺/NADH cycle that supports glycolytic ATP production and K_ATP channel closure, supporting insulin secretion (48). Increased lactate dehydrogenase isoforms, coupled with increased glucokinase expression, support the hypothesis that CIII KO islets secrete insulin normally in culture, at least partially, via glycolysis. Interestingly, lactate dehydrogenase overexpression has also been shown to be detrimental to insulin secretion in rodent models of diabetes and is upregulated in T2D human islets (49,50), supporting the discrepancy between our in vivo and ex vivo data. Other alternative hypotheses to explain this discrepancy are that there are altered neural and/or paracrine signaling pathways in the CIII KO animals that are lost when islets are examined ex vivo. It is also possible that the CIII KO isolated islets recover in culture, negating the insulin secretion defect that we observe in vivo. Additional studies will be required to better understand this discrepancy.

Taken together, our findings show that individual OxPhos complexes differentially affect pancreatic β-cell function. Specifically, that loss of CIII leads to early β-cell dysfunction and disrupted mitochondrial and cytosolic Ca²⁺ signaling. Of relevance, islets from patients with T2D have been shown to have downregulated expression of OxPhos complexes (17,18). Our findings underscore potential pathogenic roles of OxPhos deficiencies in T2D progression.

Acknowledgments. The authors thank Leah Kanakaraj (Department of Neurology, University of Miami) and Lise-Michelle Theard (Department of Neurology, University of Miami) for their assistance with mice colony maintenance and genotyping. The authors also thank Vania Almeida and the University of Miami Transmission Electron Microscopy Core for electron microscope sample preparation and assistance with the generation of electron microscope images.

Funding. This work was funded primarily by the National Institutes of Health, National Institute of Environmental Health Sciences grant R33ES025673 (A.C. and C.T.M.) and National Institute of Diabetes and Digestive and Kidney Diseases grant F32DK127691 (A.L.L.). Secondary support was provided by National Institute of Diabetes and Digestive and Kidney Diseases grants R01DK084321, R01DK111538, R01DK130328, and R01DK113093 (A.C.), National Eye Institute grant R01EY0108041, National Institute of Environmental Health Sciences grant R01NS079965, the Army Research Office (W911NF-21-1-0248), and Florida Biomedical Foundation grant 21K05 (C.T.M.).

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. A.L.L. contributed to study design, data collection, data analysis, and writing the manuscript. N.N. contributed to developing the mouse models, data collection, and data analysis. R.A.L. contributed to data collection and analysis and to editing the manuscript. A.T. contributed to islet isolations. E.P. contributed to immunohistochemical staining. C.T.M. and A.C. contributed to study design and conceptualization, provided supervision, and edited the manuscript. All authors discussed results and commented on the manuscript. C.T.M. and A.C. 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 as a poster at the 80th Scientific Sessions of the American Diabetes Association, virtual meeting, 12–16 June 2020, and as an oral abstract at the 81st Scientific Sessions of the American Diabetes Association, virtual meeting, 25–29 June 2021.