OGT Regulates Mitochondrial Biogenesis and Function via Diabetes Susceptibility Gene Pdx1

Type 2 diabetes (T2D) is the progressive state of dysregulated glucose homeostasis associated with decreased insulin sensitivity in the peripheral tissues and pancreatic β-cell failure. Improving β-cell function and preserving β-cell mass are key in the prevention of T2D. As the powerhouse of a cell, the mitochondria plays a crucial role in the maintenance of β-cell health and function by converting nutrient substrates into ATP, the energy needed to release and store insulin in response to nutrient flux, such as glucose (1). Glucose and lipids are the main fuels used by the mitochondria to generate ATP and other metabolites needed to control stimulus-secretion coupling in β-cells (2–4). Mutations in mtDNA are associated with the incidences of T2D (5,6), and mitochondrial dysfunction is linked to β-cell failure (7,8), but the mechanisms behind mitochondrial dysfunction are unclear.

The main function of the β-cell is to produce and secrete insulin in a regulated manner. Thus, the β-cells, in particular, are very sensitive to glucose flux (9,10) and other nutrients like free fatty acid (11). In mammals, the single gene encoding O-GlcNAc transferase (OGT) is located on the X chromosome. As a nutrient sensor, OGT is highly expressed in β-cells (12,13) and localized in the cytosol, nucleus, and mitochondria (14). To date, the role of OGT in the mitochondria of β-cells is unexplored.

OGT is the sole enzyme responsible for adding a posttranslational modification called O-GlcNAcylation by catalyzing the addition of a single N- GlcNAc onto serine/threonine residues of target proteins impacting biological processes, including β-cell function and survival (15). The detachment of O-GlcNAc is done by O-GlcNAcase enzyme. A single nucleotide polymorphism in MGEA5 encoding O-GlcNAcase is associated with T2D in Mexican Americans (16). OGT relies, in part, on the concentration of its substrate, UDP-GlcNAc, the product of the hexosamine biosynthetic pathway from glucose, amino acid, fatty acid, and nucleotide metabolisms (17).

Previously, we generated a tissue-specific OGT loss murine model in β-cells (βOGTKO) and showed that they develop overt diabetes by 6 weeks of age in part due to defects in β-cell mass and function (15). Hence, O-GlcNAc signaling is important for normal β-cell function, underscoring the importance of specific enzymatic targets of OGT. For example, pancreatic and duodenal homeobox 1 (Pdx1) (18,19), NeuroD1 (20), eIF4G1 (21), SERCA2 (22), and FoxO1 (23) have been shown to be direct targets of OGT in β-cells. In a transformed MIN6 cell line, high glucose increased the O-GlcNAcylation on Pdx1, thereby increasing its DNA binding activity to promote glucose-stimulated insulin secretion (GSIS) (18). Apart from being an important regulator of β-cell development and insulin synthesis (24–26), Pdx1 also controls mitochondrial function and mitophagy. Suppression of Pdx1 in rat islets in vitro (27) or heterozygous deletion of Pdx1 in adult mouse β-cells (28) resulted in mitochondrial dysfunction and impaired insulin secretion (29).

In the current study, we use genetic manipulation to ablate OGT constitutively (βOGTKO) and inducibly in mature β-cells (iβOGTKO) and investigated mitochondrial structure and function. Unexpectedly, our data reveal a temporal regulation on mitochondrial biogenesis and function by OGT in β-cells. Mechanistically, we show that mitochondrial dysfunction induced by OGT deletion in βOGTKO islets is independent of Chop-mediated endoplasmic reticulum (ER) stress but associated with reduced Pdx1 protein level. Genetic overexpression of Pdx1 increased insulin content and improved mitochondrial morphology and function in βOGTKO mice.

All procedures were performed based on the approval from University of Minnesota’s Institutional Animal Care and Use Committee (#1806–36072A). Conditional deletion of OGT in pancreatic β-cells during development (βOGTKO) is carried out by crossing mice harboring floxed Ogt gene (OGT f/f) in both alleles with mice expressing cre-recombinase under a rat insulin promoter (Rip-cre; from Dr. Pedro Herrera, University of Geneva). Conditional deletion of OGT in adult pancreatic β-cells (iβOGTKO) is carried out by crossing OGT f/f with mice expressing cre-recombinase under a mouse insulin promoter followed by tamoxifen induction (via gavage as previously described [22]) (Mip-creERTM; from Dr. Louis H. Philipson, University of Chicago). Chop^−/+ animals were purchased from The Jackson Laboratory and bred with βOGTKO to generate double-transgenic (Tg) mouse models. Overexpression of Pdx1 in β-cells lacking OGT is achieved by crossing βOGTKO and Pdx1Tg (from Dr. Yoshio Fujitani, Gunma University) mice to further generate double-Tg mouse models. All mice were group-housed on a 14:10-h light/dark cycle.

Islets were isolated from ∼4-week-old nondiabetic mice for βOGTKO. For iβOGTKO, islets were isolated at least 4 weeks after tamoxifen induction. Perfusion of collagenase solution (1 mg/mL) is carried out through common bile duct puncture, and inflated pancreas is collected and digested in a 37°C water bath for 8–13 min. Homogenate is filtered using a 70-µm cell strainer. Islets are then handpicked and cultured in a 37°C incubator with RPMI 1640 media containing 5 mmol/L glucose, 10% FBS, and 1% penicillin/streptomycin as previously described (15).

Islets were washed and prepared as previously described (21). Briefly, samples were sectioned using a diamond knife on a Leica Ultracut UCT microtome at a thickness of 70–100 μm, collected on 200-mesh copper grids, and stained with 3% aqueous uranyl acetate (20 min) and Sato’s lead citrate stain (3 min). Grids were observed and imaged on a Philips CM12 transmission electron microscope at 60 kV at the University Imaging Center (University of Minnesota). Pancreatic β-cells were located by identifying insulin granules. At least 12 cells from n = 3 animals were used for manual analysis of mitochondrial morphology for all groups except Chop^+/−;βOGTKO (n = 6 cells). To be consistent with the total number of mitochondrial structures observed within a cell, only images with clear cell boundaries were used. Mitochondria with clear loss of cristae (inner mitochondrial membrane) were considered swollen. Other normal forms (spherical and tubular) were also counted for quantification. The total number of mitochondria represent the sum of all forms (spherical, tubular, and swollen).

Islets were lysed in 1× RIPA buffer containing protease inhibitor, phosphatase inhibitor, and 1% SDS. Lysate was sonicated briefly, and the supernatant was used to measure protein concentration by bicinchoninic acid assay. An equal amount of protein was loaded onto SDS gel and transferred into a polyvinylidene difluoride membrane. The blot was blocked with 10% milk and incubated in primary antibody at 4°C for overnight; this was followed by secondary incubation with appropriate secondary antibody and subsequent imaging using enhanced chemiluminescence reagents. The following primary antibodies were used: RL2 (Abcam), specific antibody for O-GlcNAc modification, OXPHOS Cocktail (Abcam), Pdx-1 (Millipore), and vinculin (Cell Signaling Technology).

Total DNA, including genomic and mtDNA, was isolated from islets using the QIAamp DNA Micro Kit, following the manufacturer’s protocol. It was then used for quantitative real-time PCR with primers specifically flanking mouse nuclear gene, mouse β₂-microglobulin, and mouse mtDNA sequence (30). mtDNA content was calculated as the ratio of mtDNA expression over nuclear DNA expression.

Mouse islets (∼75–100) were seeded into wells of Cell-Tak–coated XFe96 plates containing 100–175 μL/well of Seahorse assay medium prepared following the manufacturer’s recommendation. Mitochondrial respiration was measured using the Seahorse XF Cell Mito Stress Test Kit for the Seahorse XFe96 Extracellular Flux Analyzer (Agilent Technologies, Santa Clara, CA). Basal respiration was first measured in 3 mmol/L glucose media. Islets were then sequentially exposed to 20 mmol/L glucose, 1 μmol/L oligomycin A, 0.5 μmol/L carbonyl cyanide-p-trifluoromethoxyphenylhydrazone (FCCP), and 0.5 μmol/L antimycin A plus rotenone. For calculation of bioenergetic parameters, average non–electron transport chain oxygen consumption rate (OCR) values after injection of antimycin A plus rotenone were subtracted from all OCR measurements (Supplementary Fig. 1). All parameters were normalized to basal respiration and reported as normalized OCR. OCR measurements were normalized to DNA concentration measured using the Quant-iT PicoGreen dsDNA Assay.

Islets from βOGTKO (4-week-old nondiabetic male mice, n = 3–4) and iβOGTKO (25-week-old male mice, 7 weeks postinduction, n = 3) mice were isolated. Islet pellets were resuspended and quantitated using Bradford assay. Total islet lysates were used for tandem mass tag labeling βOGTKO and iβOGTKO groups, respectively, along with their control samples. Samples were then analyzed on an Orbitrap Fusion mass spectrometer. The Sequest (31) HT algorithm in Proteome Discoverer v2.4 (Thermo Fisher Scientific, Waltham, MA) was used for interpretation of tandem mass spectrometry (mass spectra) and protein inference with the Percolator false discovery rate (FDR) node. Tandem mass spectrometry data from all concate nated fractions were combined into a single report (mudpit style). Search parameters were: C57BL/5J mouse universal proteome sequence database (UP000000589) from UniProt.org (19 November 2019) merged with rat OGT (A0A0G2K3V4), GFP-like chromoprotein (Q9U6Y5), and common laboratory contaminant proteins from the global proteome machine at https://www.thegpm.org/crap/. The spectral mass recalibration was invoked with 20-ppm precursor tolerance and 0.1-Da fragment ion tolerance, TMT6plex (+229.16293) variable modification on lysine and protein N terminus, and methylthio modification of cysteine (+45.98772). The SEQUEST HT search parameters were: precursor mass error tolerance, 15.0 ppm; fragment mass error tolerance, 0.1 Da; precursor mass search type monoisotopic; enzyme full trypsin peptides, maximum two missed cleavage sites; variable modifications methionine oxidation (+15.994915), pyroglutamic acid (−17.026549), deamidation of asparagine and glutamine (+0.984016), protein N-terminal acetylation (+42.0 10565), TMT6plex on lysine and protein N terminus; fixed modification methylthio modification of cysteine; and maximum dynamic variable modifications per peptide 3. We set the protein FDR validator strict target to 0.01 FDR and the relaxed target to 0.05 FDR.

We used Proteome Discoverer 2.4 for relative protein quantification with the following parameter settings in the Reporter Ion Quantifier node: unique and razor peptides were included, shared peptides were excluded from quant measurements, reporter ion correction factors were applied, the precursor coisolation threshold was 50, and the average reporter ion S/N threshold was 10. We applied total peptide amount for normalization; we set the protein ratio calculation to pairwise peptide ratio-based and the hypothesis test to Student t test; P values were adjusted using Benjamini-Hochberg correction for FDR. Volcano plots were generated using the Benjamini-Hochberg adjusted P values and abundance ratios estimated by Proteome Discoverer. All plots were generated using the R statistical programming language and the R package ggplot2 (https://ggplot2.tidyverse.org).

Mice were fasted for 16 h prior to the test. A total of 2 g/kg of mouse body weight of 50% dextrose solution was injected intraperitoneally. Blood glucose was measured using glucometer after 30, 60, and 120 min postinjection. In vitro insulin secretion in response to glucose or palmitate was done as previously described (22).

Post-GSIS/fatty acid–stimulated insulin secretion (FASIS) islet screens or naive islet pellets were sonicated, in a cold room or on ice, in lysis buffer as previously described (22). ALPCO Mouse Ultrasensitive Insulin ELISA (80-INSMU-E01) and Mouse Proinsulin ELISA (80-PINMS-E01) were used for insulin content and proinsulin content measurement, respectively. Samples for insulin assessment were diluted in ELISA wash buffer and for proinsulin assessment in ELISA zero standard solution to fit the standard curve, and values were measured following the manufacturer’s protocol. Insulin content from post-GSIS islets includes both secreted insulin and islet insulin content as determined at the end of the test. Total DNA used to normalize were determined from lysates using a PicoGreen dsDNA assay, per kit instructions. Fluorescence (480/520 nm) was read on the BioTek plate reader and calculated by a linear regression analysis in Excel of an in-plate stock dilution ladder.

Data are represented as mean ± SEM. For graphs with a single time point, an unpaired t test with Welch correction was performed. Repeated-measures two-way ANOVA with Bonferroni multiple correction was used to analyze graphs with multiple time points. For grouped analysis, repeated-measures two-way ANOVA with Sidak multiple correction was used. Results were analyzed and plotted using Prism 8 (GraphPad Software, La Jolla, CA) and considered statistically significant when the P value was <0.05.

The data sets generated during and/or analyzed during the current study are available from the corresponding author upon reasonable request.

OGT is ubiquitously expressed in mammalian cells; however, it is highly expressed in the pancreas (12). To date, however, the role of OGT in β-cell mitochondria remains unexplored. After successful validation of OGT activity reduction in the conditional and constitutive model βOGTKO (Fig. 1A and B), we characterized the morphology of the mitochondria in 1-month-old normoglycemic βOGTKO mice (Fig. 1C). Residual O-GlcNAc signal, as measured with the RL2 antibody, is contributed by the extra–β-cell (i.e., α- and δ-cells) population present in whole islets. Electron microscopic (EM) images of β-cells revealed structural changes in the mitochondrial morphology in βOGTKO compared with control mice (Fig. 1D and E). We observed a significant increase in the total number of swollen mitochondria with apparent disruption of cristae in the inner mitochondrial membrane (white arrow in the inset of Fig. 1E) in βOGTKO mice. Moreover, a decrease in the number of fragmented (spherical) and tubular mitochondria was evident (Fig. 1D–F). No alteration in the total number of mitochondria per β-cell was observed in βOGTKO compared with control mice (Fig. 1F), and this finding was corroborated by the equivalent level of mtDNA content (Fig. 1G). These data suggest that OGT plays an important role in maintaining normal morphology of the mitochondria.

Next, we tested whether ablation of OGT impacted mitochondrial function in young and normoglycemic βOGTKO mice. Islets were isolated from 1-month-old βOGTKO mice, and their mitochondrial bioenergetics were studied using the Seahorse XFe96 Analyzer ex vivo (Fig. 2A). Islets from βOGTKO mice exhibited higher basal respiration compared with control (Fig. 2B); therefore, the overall OCR were normalized to their respective basal respiration. The schematic diagram for analyzing glucose stimulation, ATP production, proton leak, and maximum respiration are shown in Supplementary Fig. 1A. There were significant impairments in 20 mmol/L glucose-stimulated cellular respiration and ATP production in the islets of βOGTKO mice (Fig. 2C and D) without any alterations in maximal respiration (Fig. 2E) and proton leak (Fig. 2F). There was also a significant reduction in anaerobic glycolysis observed under high-glucose conditions (20 mmol/L) in islets from βOGTKO mice (Fig. 2G and H). These data suggest that deletion of OGT disrupts cellular bioenergetics via its effect on both mitochondrial function and glycolysis.

To characterize the role of OGT in the mitochondria of mature β-cells, we deleted OGT in 12-week-old mice. We validated OGT reduction of activity via RL2, a specific O-GlcNAc antibody, in islets of the inducible model, iβOGTKO (Fig. 3A and B). Prior to the collection of islets for EM, we validated that OGT loss in mature β-cells does not cause overt diabetes (Fig. 3C), but glucose intolerance in vivo was previously reported (15) (Fig. 3D and E). Next, we evaluated β-cells in iβOGTKO mice by EM (Fig. 3F and G). Quantitative analysis of mitochondrial morphology revealed increased number of mitochondria with spherical and tubular morphology with significant increase in the total number of mitochondria per β-cell in iβOGTKO compared with control mice (Fig. 3H). Moreover, quantification of mtDNA showed increased mitochondrial copy number in the islets of iβOGTKO mice (Fig. 3I).

Next, we assessed mitochondrial function in islets of iβOGTKO mice. Although the total number of mitochondria was increased at the single-cell level and the basal respiration was reduced in iβOGTKO islets, the normalized OCR (corrected by total DNA) over the basal respiration was not altered in whole intact islets between the iβOGTKO and control mice (Fig. 4A and B). There were no differences in glucose-stimulated OCR and ATP production observed (Fig. 4C and D), and iβOGTKO islets showed similar levels of maximal respiration rate (Fig. 4E) and proton leak (Fig. 4F). The extracellular acidification rate (ECAR) of live cells (Fig. 4G) was increased in iβOGTKO islets compared with controls (Fig. 4H), indicating improved glycolysis rate under high-glucose (20 mmol/L) conditions. Together, these findings suggest that ablation of OGT in mature β-cells does not impair ATP production, which may be explained in part by the increased number of mitochondria per cell.

It is well known that ER stress can disrupt mitochondrial function, leading to apoptosis (32–35). Islets from βOGTKO mice have been shown previously to have increased ER stress, and reduction of Chop in βOGTKO improves hyperglycemia and β-cell mass in Chop^+/−;βOGTKO mice (15). Hence, we tested whether ER stress can contribute to the observed decrease in mitochondrial function in βOGTKO mice. We tested whether relieving Chop-dependent ER stress improves mitochondrial function in islets with β-cell OGT deletion. EM images of β-cells in Chop^+/−;βOGTKO mice displayed increased number of swollen mitochondria with apparent disruption of the inner mitochondrial membrane (cristae) and decreased number of normal mitochondria (Fig. 5A and B), which was similar to the mitochondrial morphology observed in the βOGTKO mice (Fig. 1D–F). Islets from Chop^−/−;βOGTKO mice also displayed significant impairment in glucose-stimulated cellular respiration (Fig. 5C and D) and ATP production (Fig. 5E). When compared with βOGTKO islets, Chop^−/−;βOGTKO islets showed comparable maximal respiration rate (Fig. 5F) and proton leak (Fig. 5G). In contrast to βOGTKO islets, Chop^−/−;βOGTKO islets did not show any differences in glycolysis (Fig. 5H). Taken together, ablation of Chop-mediated ER stress does not improve mitochondrial morphology and function in βOGTKO mice, suggesting that OGT regulates mitochondrial morphology and function independent of Chop-mediated ER stress.

To further characterize the role of OGT in the mitochondria, we assessed the protein subunits of oxidative phosphorylation and their relative expression levels (Supplementary Fig. 2). Using Western blotting, we observed comparable levels of complex V, IV, III, and II between islet lysates from βOGTKO and littermate controls (Supplementary Fig. 2A–E). However, in islets from iβOGTKO mice, we observed a significant increase in protein levels of ATP-5A (complex V) and UQCRC2 (complex III) in iβOGTKO islets (Supplementary Fig. 2F–H). Complex II and IV were comparable between islet lysates from iβOGTKO and littermates (Supplementary Fig. 2F, I, and J).

In order to uncover possible mechanisms by which OGT may regulate mitochondrial morphology and function in β-cells, we performed an unbiased and quantitative proteomic analysis in islets isolated from 1-month-old βOGTKO and at 1 month post–tamoxifen induction (at 3 months old) in iβOGTKO mice. We identified 81 and 100 proteins differentially expressed with a significant P value of <0.05 in βOGTKO and iβOGTKO islets, respectively, when compared with their corresponding controls (Fig. 6A). There are 26 proteins differentially expressed in both models of OGT loss (Fig. 6A). Consensus analyses show that specific proteins (e.g., Aass, Tspan6, and Aldob) were altered in the same direction in both models (Fig. 6B). The volcano plots in Fig. 6C and D show proteins altered in βOGTKO and iβOGTKO, respectively. In both models, GFP, the reporter used for Cre lineage, was detected, and this served as a positive control for Cre activation (Fig. 6C and D). Zinc transporter 8 (Slc30a8) and hemoglobin subunit β-2 (Hbb-b2) were reduced, whereas calmodulin regulator protein (Pcp4), argininosuccinate synthase 1 (Ass1), and aminoadipate-semialdehyde synthase (Aass) were among the increased protein in islets of βOGTKO compared with controls. In iβOGTKO islets, in addition to OGT, histocompatibility 2 (H2-K1), prodynorphin (Pdyn), phosphoglycerate dehydrogenase (Phgdh), and UDP-glucuronosyl transferase (Ugt2b34) were among the proteins reduced compared with controls. Matrilin 2 (Matn2), surfactant protein D (Stfpd), aldolase fructose-bisphosphate B (Aldob), isovaleryl-CoA dehydrogenase (Ivd), and tetraspanin 6 (Tspan6) were the top five proteins increased in iβOGTKO. Next, Ingenuity Pathway Analysis was carried out to assess top canonical pathways, networks, and upstream regulators in the proteome of either βOGTKO and iβOGTKO (Fig. 7A). The top canonical pathway in βOGTKO was maturity-onset diabetes of the young (MODY), and the top upstream regulator and second top network was Pdx1, a MODY type 4 (Fig. 7B). The glucose network in both models is shown in Fig. 7A. The top canonical pathway in the iβOGTKO proteome was lipopolysaccharide/interleukin-1–mediated inhibition of retinoid X receptor (RXR) function, and CLN3 (lysosomal/endosomal transmembrane protein) was the top upstream regulator and causal pathway (Fig. 7B). The second upstream regulator in iβOGTKO was HNF1A (Fig. 7B).

The most significant upstream regulator in the βOGTKO transcriptome and proteome analysis was Pdx-1, a β-cell OGT target with major roles in defining β-cell identity and function (36,37). As a master transcription factor in β-cells, Pdx1 has also been shown to regulate mitochondrial morphology, biogenesis, and function (27–29). Additionally, it was shown that Pdx1 deficiency causes mitochondrial dysfunction through suppression of TFAM (28). Therefore, we postulated that Pdx1 protein was reduced, and indeed, we observed a significant reduction in Pdx1 levels in βOGTKO islets (Fig. 8A and B). In contrast, islets isolated from iβOGTKO mice with normal mitochondrial morphology and function showed normal Pdx1 protein levels (Fig. 8C and D). Therefore, we tested whether genetic reconstitution Pdx1 in β-cells (Rip-cre; CAG-CAT-Pdx1 [38], Pdx1Tg) can rescue mitochondrial dysmorphology in the βOGTKO mice. By EM analysis, we observed an increased in spherical and tubular-shaped mitochondria in the βOGTKO;Pdx1Tg compared with littermate βOGTKO mice (Fig. 8E–H). As previously demonstrated in Fig. 1D, OGT loss increased the frequency of swollen-like mitochondria, and genetic overexpression of Pdx1 normalized the number to the control level (Fig. 8H). The total number of mitochondria was also increased in βOGTKO;Pdx1Tg compared with littermate βOGTKO or controls (Fig. 8H). These data suggest that reconstitution of Pdx1 restores mitochondrial morphology in βOGTKO islets.

Next, we investigated whether Pdx1 overexpression rescues mitochondrial dysfunction in the βOGTKO islets (Fig. 9A). The cellular respiration in response to glucose was a trend toward reduction (P = 0.06) between βOGTKO and littermate controls; however, a comparable response was observed between βOGTKO;Pdx1Tg and control mice (Fig. 9B). A significant improvement in maximal respiration (Fig. 9C) was observed in βOGTKO;Pdx1Tg islets compared with βOGTKO islets. Comparable levels of ATP and proton leak were observed among genotypes tested (Fig. 9D and E). The ECAR as a result of glycolysis was reduced in βOGTKO compared with controls, and the overexpression of Pdx1 normalized it to control levels (Fig. 9E and F). These data suggest that reconstitution of Pdx1 improves glycolysis and mitochondrial respiration.

Next, we tested whether overexpressing Pdx1 in β-cells in conjunction with βOGTKO can rescue the in vitro deficits in insulin content, proinsulin processing, and FASIS, three hallmark phenotypes of islets βOGTKO mice (15,22,39). We previously reported reduced insulin and proinsulin content in islets from βOGTKO mice (15,39). We found that islets from βOGTKO;Pdx1Tg mice increased both insulin and proinsulin content compared with βOGTKO islets (Supplementary Fig. 3A and B), consistent with Pdx1’s role as a major insulin biogenesis transcription factor. However, proinsulin processing (measured by the ratio of proinsulin to mature insulin levels) was not rescued as islet proinsulin/insulin ratios in the βOGTKO;Pdx1Tg islets were the same compared with βOGTKO islets (genotype P = 0.3 by two-way ANOVA) (Supplementary Fig. 3C). We previously reported that GSIS, as a percent of content, was not altered between control and βOGTKO islets, but OGT loss prevented the expected increase in GSIS in the presence of acute palmitate (FASIS) (22). We reiterated those results in this study, showing no palmitate-induced differences in basal (low glucose vs. low glucose in the presence of 100 μmol/L palmitate) or glucose-stimulated (high glucose vs. high glucose or the same in the presence of 100 μmol/L palmitate) insulin release in βOGTKO islets nor in βOGTKO;Pdx1Tg islets, indicating a failure of this model to rescue FASIS (Fig. 9G).

The role of OGT in mitochondrial structure and function has not been investigated previously in pancreatic β-cells. By genetic manipulation, we show that OGT plays a crucial role in maintaining β-cell mitochondrial structure and function. Mechanistically, we identified that the mitochondrial dysfunction in OGT-deficient β-cells was associated with reduced Pdx1 protein levels. We show that genetic reconstitution of Pdx1 in β-cells, but not ablation of Chop-mediated ER stress, was sufficient to rescue mitochondrial morphology and dysfunction and also improved insulin content in islets with OGT loss.

Mitochondrial dysfunction and alterations in morphology have been reported in diabetic animal models (40–43) and patients with T2D (44). In both humans and mice, under diabetic conditions, swelling of the mitochondria may occur and is accompanied with compromised function (8,45,46), and a genetic linkage has been established between mitochondrial DNA alterations and T2D (47). In this study, we report abnormal mitochondrial morphology and dysfunction in β-cells of 1-month-old and normoglycemic in βOGTKO mice. These mice become diabetic by 2 to 3 months of age, in part due to sustained loss of insulin content and β-cell mass failure (Fig. 9H). In the inducible model, in which OGT is genetically deleted at 3 months of age on, we observed preserved mitochondrial structure, although they are glucose intolerant and display insulin secretion defect without loss of β-cell mass (15). The health of the mitochondria is closely linked to β-cell function. As a nutrient sensor downstream of glucose, lipids, and amino acids, OGT offers a mechanistic link between nutrient stimuli and mitochondrial function in β-cells. Our study provides the first evidence to show that OGT regulates mitochondrial morphology and function in β-cells, and the outcome of OGT ablation is distinct when deleted in either the embryonic stage or in mature β-cells.

Deletion of OGT constitutively in β-cells resulted in abnormal mitochondrial morphology and severe dysfunction (reduced OCR and ATP production) in 1-month-old mice. These data are supported by previous findings showing reduced OCR in OGT deletion in the liver (48) and muscle (49). In contrast, ablation of OGT in mature β-cells of 3-month-old mice does not grossly impair OCR and ATP production. This may be explained in part by a compensation of increased numbers of mitochondria in β-cells of iβOGTKO mice. In C2C12 myoblasts, Wang et al. (50) reported that increased mitochondrial O-GlcNAcylation promoted PGC-1α degradation, resulting in lower mitochondrial biogenesis. Thus, the increased mitochondrial biogenesis seen in iβOGTKO islets may involve an increase in the stability of PGC-1α. Gawlowski et al. (51) reported that increased O-GlcNAcylation augments the translocation and activity of DRP1 to promote mitochondrial fission. Thus, ablation of OGT would lead to reduced DRP1 activity and thereby more fused mitochondria (51,52). In the iβOGTKO islets, we observed increased tubular and spherical mitochondria and increased mtDNA content concomitant with normal mitochondrial function under high-glucose conditions but with lower OCR under basal conditions. The reduction in basal OCR and increased glycolysis observed in iβOGTKO islets have also been observed in OGT-deleted skeletal muscle (49), HeLa cells (53), and neonatal rat cardiomyocytes (42,43).

There are possible explanations why there are distinct mitochondrial phenotypes between the βOGTKO and the inducible OGT loss models. For example, in β-cells of βOGTKO mice, it is likely that increased diabetic condition (increased insulin demand, hyperglycemia, and ER stress) contributed to the phenotypes observed. Indeed, mitochondrial function can be regulated by O-GlcNAcylation, particularly under diabetic circumstances (54). Although both models presented normoglycemic at the time of islet harvest (based on a one-time-point assessment of nonfasting glucose), at the molecular level, only βOGTKO islets demonstrated a significant reduction in insulin biogenesis (reduced Ins and Pdx1 mRNA and protein levels), altered calcium signaling, and elevated ER stress (15,39). The temporal regulatory role of OGT in mitochondrial function may also arise in part due to the sustained loss of O-GlcNAcylation on specific OGT targets, such as mitochondrial protein DRP1 (51), Milton (55), Pdx1 (18), and TFAM (28,56) during early development or particularly in the intrauterine environment and whether these mitochondrial regulator proteins are significantly altered to impact mitochondrial biogenesis, dynamics, and function.

The mitochondria and ER form structural and functional networks, and they cooperate closely to regulate cell survival. We previously demonstrated that ablation of Chop-mediated ER stress was sufficient to lower nonfasted glucose levels and improve β-cell mass in βOGTKO mice (15). However, in the current study, ablation of Chop-mediated ER stress was insufficient to rescue mitochondrial dysmorphology or dysfunction in islets of βOGTKO mice. Therefore, it is possible that non-CHOP–dependent ER stress pathways may have contributed to the defect in mitochondrial morphology and function observed in βOGTKO islets.

The proteomes of βOGTKO and iβOGTKO islets were distinct and with limited number of overlapping proteins changed in the same direction. Because mitochondrial defect was only seen in βOGTKO islets, we reasoned to follow up on nonoverlapping pathways. We were intrigued that MODY signaling rose to the top of pathways identified in βOGTKO islets. Moreover, the homeodomain transcription factor Pdx1, a gene associated with both T2D and a MODY type 4, as well as key regulator of mitochondrial function and mitophagy (27–29), was identified as top upstream regulator of altered proteins in the βOGTKO proteome and transcriptome (22). Indeed, we identified different protein levels of Pdx1 in βOGTKO versus iβOGTKO islets, but this is only one of the many possible explanations for the distinct mitochondrial phenotypes between these mice. Pdx1 was previously shown to be modified by OGT (37), and its O-GlcNAcylation affects its DNA binding activity (18). The sustained Pdx1 loss in βOGTKO may result in reduced glucose transporter Glut2 (57) and thereby glucose level utilization in the cell. This may explain the glycolysis differences between the two models and may indicate a change in the supply of pyruvate available to fuel mitochondrial oxidative phosphorylation and the potential for elevated glycolytic flux in the inducible model to partially compensate for mitochondria-level deficits in glucose/pyruvate utilization. A future direction is to explore the role of other possible OGT targets independent of Pdx1, such as DRP, TFAM, and PGC-1α in β-cells.

In summary, we showed that genetic reconstitution of Pdx1 reversed the morphological phenotypes (e.g., swelling) as well as mitochondrial respiration and glycolysis in βOGTKO islets. Collectively, our data suggest that distinct proteins may contribute to specific defects observed in islets with loss of OGT. In summary, the regulation of OGT in mitochondrial morphology and function highlights the importance of OGT in cell biology and may serve as a link between nutrient signal and mitochondrial function to promote normal β-cell physiology.

Acknowledgments. The authors thank Dr. David Bernlohr (University of Minnesota) for discussion; Minnesota Supercomputing Institute for an Updraft Grant, which supported the analysis of the proteomics data; Drs. Leann Higgins and Todd Markowski at the University of Minnesota Proteomic Core and Gail Celio from the University Imaging Center; and Brian Akhaphong and Alicia Wong (University of Minnesota) for technical support.

Funding. This work was supported by the National Institutes of Health (National Institute of Diabetes and Digestive and Kidney Diseases grants R21DK112144 and R01 DK115720 to E.U.A., grant R01 DK084236 to E.B.-M., and National Institutes of Health/National Eye Institute grants R01EY028554 and R01EY026012 and the Lindsay Family Foundation to D.A.F.).

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

Author Contributions. R.M. and E.U.A. designed experiments, analyzed data, interpreted the data, and wrote and edited the final manuscript. R.M., S.J., A.L., K.M., and A.E. generated and analyzed data and edited the final manuscript. D.A.F. and Y.F. provided resources and edited the final manuscript. E.B.-M. provided resources and contributed to the discussion. E.U.A. conceived the study, designed experiments, interpreted the data, wrote and edited the manuscript, and acquired funding. E.U.A. is the guarantor of this work and, as such, had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Prior Presentation. This study was presented at the 78th Scientific Sessions of the American Diabetes Association, Orlando, FL, 22–26 June 2018.