Comparative Proteomics Reveals Dysregulated Mitochondrial O-GlcNAcylation in Diabetic Hearts

Abstract

O-linked β-N-acetylglucosamine (O-GlcNAc), a post-translational modification on serine and threonine residues of many proteins, plays crucial regulatory roles in diverse biological events. As a nutrient sensor, O-GlcNAc modification (O-GlcNAcylation) on nuclear and cytoplasmic proteins underlies the pathology of diabetic complications including cardiomyopathy. However, mitochondrial O-GlcNAcylation, especially in response to chronic hyperglycemia in diabetes, has been poorly explored. We performed a comparative O-GlcNAc profiling of mitochondria from control and streptozotocin (STZ)-induced diabetic rat hearts by using an improved β-elimination/Michael addition with isotopic DTT reagents (BEMAD) followed by tandem mass spectrometric analysis. In total, 86 mitochondrial proteins, involved in diverse pathways, were O-GlcNAcylated. Among them, many proteins have site-specific alterations in O-GlcNAcylation in response to diabetes, which suggests that protein O-GlcNAcylation is a novel layer of regulation mediating adaptive changes in mitochondrial metabolism during the progression of diabetic cardiomyopathy.

Graphical Abstract

■ INTRODUCTION

Since its discovery over 30 years ago,^1,2 the cycling of O-linked β-N-acetylglucosamine (O-GlcNAc) on serine and threonine residues of proteins (O-GlcNAcylation) has been shown to play crucial roles in diverse biological processes. The highly dynamic O-GlcNAc modification (O-GlcNAcylation) is regulated by two enzymes: O-GlcNAc transferase (OGT) and O-GlcNAcase, which add and remove the O-GlcNAc moiety, respectively. As a major nutrient sensor, protein O-GlcNAcylation, on myriad proteins, underlies the progression of several metabolic diseases including diabetes and its complications.^3,4 Of note, dysregulated O-GlcNAcylation of proteins is a key factor in the etiology of cardiomyopathy.^5–9

Mitochondria are at the center stage of regulating cellular metabolism, and mitochondrial dysfunction underlies the progression of cardiovascular disease.¹⁰ Changes in protein abundance as well as in post-translational modifications on mitochondrial proteins have been revealed as critical to altered mitochondrial function.^11–13 However, few studies have been focused on protein O-GlcNAcylation in mitochondria especially in functional cardiac mitochondria, although a number of techniques have been developed for probing protein O-GlcNAcylation in cardiovascular and other biological settings.^14–23 One reason is that mitochondria used to be regarded as an organelle with few O-GlcNAcylated proteins.^2,24 Accumulating evidence in recent years shows that protein O-GlcNAcylation does exist in mitochondria.^25–27 An overall increased mitochondrial O-GlcNAcylation in neonatal cardiomyocytes exposed to high glucose was observed.²⁵ The very recent identification of mitochondria O-GlcNAc cycling components (i.e., OGT, O-GlcNAcase, and a UDP-GlcNAc transporter), the altered enzymatic activity of mitochondrial OGT and O-GlcNAcase, and the mislocalization of OGT suggest that mitochondrial O-GlcNAcylation plays a role in the progression of diabetic cardiomyopathy.²⁶ On the other hand, rats acutely treated with Thiamet G, a specific inhibitor of O-GlcNAcase, show an overall increased O-GlcNAcylation on mitochondrial proteins, concomitant with enhanced mitochondrial bioenergetics.²⁷ Thus, it is fundamentally important to determine how the mitochondrial O-GlcNAcome responds to a chronic disease, such as diabetes, and how the pattern of O-GlcNAcylation may differ from changes after acute treatment.

Herein, by coupling isotopic labeling and refined O-GlcNAc enrichment with tandem mass spectrometry techniques, we performed global profiling of O-GlcNAcylation in functional mitochondria from normal or STZ-induced diabetic hearts. We found that the cardiac mitochondrial proteome and O-GlcNAcome were substantially different in the two groups. Of note, a number of metabolic enzymes show altered protein abundance as well as site-specific O-GlcNAcylation. Our results suggest that mitochondrial protein O-GlcNAcylation could serve as a critical factor in the initiation and progression of diabetic cardiomyopathy.

■ EXPERIMENTAL PROCEDURES

Experimental Animals

Male Sprague–Dawley rats (8–10 weeks) were randomly split into two groups. Control rats received an injection of vehicle buffer. Diabetic rats were treated as described previously.²⁸ Briefly, rats were intraperitoneally injected with a single dose of 75 mg/kg Streptozocin (Sigma). Rats with high blood glucose level (>600 mg/dL) were housed for 4 weeks and sacrificed by decapitation. All animal experiments were approved by the Institutional Animal Care and Use Committee of the Johns Hopkins University.

Mitochondria Isolation

Mitochondria were isolated by differential centrifugation²⁹ with modifications to obtain highly purified mitochondrial pellets using Percoll-gradient ultracentrifugation. In brief, rat hearts were homogenized in ice-cold Isolation Buffer (75 mM sucrose, 225 mM mannitol, 1 mM EGTA, pH 7.4, supplemented with 1× Protease Inhibitor Cocktail 1 (Roche) and 2 μM O-(2-Acetamido-2-deoxy-D-gluco-pyranosylidene) amino N-phenyl carbamate (Sigma)) using a Potter-Elvehjem glass homogenizer. Then fatty acid-free bovine serum albumin (BSA, final conc.: 0.2%) was immediately added, with the resulting homogenates centrifuged for 5 min at 500 × g. The resulting supernatants were centrifuged for 10 min at 7700 × g, with pellets washed twice thereafter by centrifugation at 7700 × g for 5 min each. The isolated mitochondria were used for functional assays. For the purity assay and proteomics experiments, mitochondrial pellets were resuspended in isolation buffer A and layered onto a discontinuous gradient consisting of (from top to bottom) 6% Percoll, 17% and 35% Histodenz (Sigma), each made up in sucrose buffer. After centrifugation with a Ti41 rotor (Beckman) at 17 500 rpm for 30 min, the layer at the 17%/35% interface was collected and diluted with Isolation Buffer and centrifuged at 10000 × g to remove Histodenz. The protein concentration was then determined by using the BCA Assay kit (Thermo Fisher Scientific).

Mitochondrial Functional Assessment

Mitochondrial oxygen consumption rates were determined with a 96-well extracellular flux analyzer (XF96; Seahorse Bioscience), as described previously.³⁰ In brief, 5–15 μg of isolated mitochondria in assay buffer (137 mM KCl, 2 mM KH₂PO₄, 0.5 mM EGTA, 2.5 mM MgCl₂, 20 mM HEPES at pH 7.2, and 0.2% fatty acid free BSA) was transferred into a 96-well XF96 plate precoated with polyethylenimine. The plate was spun at 3000 × g for 7 min at 4 °C and then incubated at 37 °C for 20 min. The respiration was measured with the flux analyzer in the presence of 5 mM glutamate/malate [G/M] or 5 mM succinate without 1 mM ADP (State 4) or with 1 mM ADP (State 3). Mitochondrial Ca²⁺ uptake capacity was assayed as reported previously,³¹ with mitochondria (0.5 mg) suspended in a potassium-based buffer (i.e., 137 mM KCl, 2 mM KH₂PO₄, 20 μM EGTA, 20 mM HEPES, and 5 mM glutamate/malate at pH 7.15). Calcium green-5N fluorescence was recorded by a fluorometer (PTI Quantamaster) at 37 °C (excitation, 505 nm; emission, 535 nm), and CaCl₂ was added sequentially (15 μM free Ca²⁺ for the first addition and 25 μM for subsequent additions). NADH fluorescence was recorded either in the presence of 5 mM G/M or 5 mM succinate (excitation, 350 nm; emission, 450 nm). Tetramethylrhodamine methyl ester was used for ratiometric measurement of mitochondrial membrane potential (ΔΨm) by monitoring emission at 590 nm with excitations of 546 and 573 nm, as described previously.³¹

Mitochondrial Sample Purity Assay

The purity of the rat heart mitochondrial preparation was assessed according to Foster et al.²⁹ Briefly, equal amounts of proteins from the postnuclear supernatant/cytosolic fraction and Percoll-gradient purified mitochondria were supplemented with 4× Laemmli buffer (Bio-Rad). After being loaded onto the top of a 10% sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) gel (Bio-Rad) overlaid with 0.5% agarose, proteins were separated at 110 V for 1.5 h. Proteins in gels were electrotransferred to a poly(vinylidene difluoride) (PVDF) membrane. The PVDF membrane was blocked using 5% BSA in 1× TBST (i.e., 100 mM Tris-HCl, 0.9% (w/v) NaCl, 0.1% (v/v) Tween-20)) and incubated with antibodies to SDHA (Invitrogen), NDUFS3 (Invitrogen), ATP5B (Invitrogen), alpha-tubulin (Sigma), or GAPDH (Sigma). After overnight agitation, the membrane was then incubated with specific secondary horseradish peroxidase (HRP)-conjugated antibodies (Santa Cruz) under gentle agitation for 1 h. Chemiluminescence was performed under manufacturer protocols by using ECL films (GE Healthcare). Blots were washed five times for 10 min in 1× TBST between blocking, antibody incubations, and chemiluminescence detection.

Two-Dimensional Gel (IEF/SDS-PAGE) Analysis of Mitochondrial Samples

The highly purified mitochondrial pellets were lysed in buffer containing 50 mM Tris-HCl (pH 7.5), 150 mM NaCl, 1% Triton X-100, 1 mM NaF, 1 mM β-glycerophosphate, 2 μM O-(2-Acetamido-2-deoxy-D-glucopyranosylidene) amino N-phenyl carbamate, and 1× Protease Inhibitor Cocktail 1 (Sigma). After being sonicated for 2 × 10 s, the suspension was spun down at 12 000 × g for 15 min. The supernatant was collected, and the mitochondrial proteins were extracted by acetone precipitation. The precipitated proteins were vacuum-dried and then dissolved in Rehydration Buffer (8 M urea, 2% CHAPS, 50 mM DTT, 0.2% Ampholytes (Bio-Rad)). After being immersed in the protein suspensions overnight, Readystrip IPG strips (Bio-Rad) were transferred onto an Ettan IPGphor3 IEF system (GE Healthcare) for IEF separation following the manufacturer instructions. The strips were incubated sequentially with DTT-Equilibration Buffer (containing 50 mM Tris-HCl, pH 8.8, 6 M urea, 30% glycerol, 2% SDS, 0.002% Bromophenol Blue, and 65 mM DTT) and Iodoacetamide-Equilibration Buffer (containing 50 mM Tris-HCl, pH 8.8, 6 M urea, 30% glycerol, 2% SDS, 0.002% Bromophenol Blue, and 135 mM iodoacetamide) for 30 min. The strips were then loaded onto 10–20% Criterion Gels (Bio-Rad) for a second dimensional SDS-PAGE separation. Proteins in gels were electrotransferred to a PVDF membrane and probed with self-purified CTD110.6 and RL2 antibody (using mouse IgM-HRP and mouse HRP-conjugated IgG as secondary antibody (Santa Cruz), respectively) with the same procedures mentioned above. Primary antibodies incubated with 1 M GlcNAc for at least 2 h at 4 °C before blotting were used for competitive assay.

Mitochondrial Sample Preparation and TMT Labeling

Acetone precipitated mitochondrial proteins from four control rats and four diabetic rats were vacuum-dried and dissolved in 8 M urea supplemented with 50 mM triethylammonium bicarbonate (pH 8), with the protein concentration determined by the BCA assay. An equal amount of proteins (9 mg of each) was reduced with 10 mM DTT, alkylated with 30 mM iodoacetamide, and in-solution digested overnight by trypsin. The resulting peptides were then treated with calf alkaline phosphatase (New England Biolab) and PNGase F (New England Biolab) for another 12 h. The peptides were desalted with C18 Sep-Pak columns (Waters) and dried with lyophilization. A small portion (1%) of the peptides was isotopically labeled with a 6-plex TMT kit (Thermo Fisher Scientific) according to the manufacturer’s protocol. The labeled peptides were pooled, fractionated by a basic pH reversed-phase liquid chromatography (bRPLC, as described in a previous report³²) into 12 fractions, dried down, and then analyzed by liquid chromatography–tandem mass spectrometry (LC–MS/MS) for protein quantification.

O-GlcNAc Peptide Enrichment

The remaining peptides (99%) from control and diabetic cardiac mitochondria were subjected to beta-elimination and Michael-addition with d0- and d6-DTT (BEMAD), respectively, with slight modifications.^33–35 Briefly, peptides were dissolved in BEMAD buffer (1.5% (v/v) triethylamine, 20 mM d0-DTT or d6-DTT, 20% (v/v) ethanol, pH 12.5 by NaOH). After incubation at 50 °C for 4 h, the reaction was neutralized with trifluoroacetic acid. The DTT-substituted peptides were pooled and incubated with thiol-sepharose resin (Sigma) in degassed PBS containing 1 mM EDTA (PBS/EDTA, pH 7.4) for 4 h. The resin was extensively washed with PBS/EDTA and 40% acetonitrile, and then DTT-containing peptides were released by PBS/EDTA containing 20 mM free DTT. Eluted peptides were desalted with C18 spin columns (Nest Group) for LC–MS/MS analysis.

LC–MS/MS Analysis

An LTQ-Orbitrap Velos (Thermo Fisher Scientific) attached to Waters UPLC chromatography system was used for MS/MS analysis of the peptides. Loaded peptides were trapped on a 2 cm trap column (YMC gel ODS-A S-10 μm), fractionated with a 75 μm × 15 cm column (Magic C18 AQ, 5 μm, 100 Å (Michrom Bioresources)), and then electrosprayed through a 15-μm emitter (New Objective). A reversed-phase solvent gradient, which consisted of solvent A (0.1% formic acid) with increasing levels of solvent B (0.1% formic acid, 90% acetonitrile) over a period of 90 min, was used for peptide separation. The LTQ Orbitrap Velos was set at 2.0 kV spray voltage, full MS survey scan range m/z 350–1800 m/z, data-dependent HCD MS/MS analysis of top eight precursors with minimum signal of 2000, 30 s dynamic exclusion limit, isolation width of 1.2, and normalized collision energy of 40. The precursor and fragment ions were analyzed at resolutions 30 000 and 15 000, respectively.

Database Search and Data Interpretation

MS/MS spectra were processed by Proteome Discoverer (v1.4 Thermo Fisher Scientific) using three nodes: common, Xtract (spectra were extracted, charge state deconvoluted, and deisotoped using Xtract option), and MS2 Processor. MS/MS spectra from three nodes were then analyzed with Mascot v.2.2.2 (MatrixScience) using the RefSeq rat database (2012), with concatenated decoy database, specifying Rattus as species, trypsin/P as enzyme, missed cleavage 2, precursor mass tolerance 15 ppm, and fragment mass tolerance 0.03 Da. For TMT-labeled samples, carboamidomethylation (Cys) and TMT (Lys and N-term) were set as fixed modifications, and deamidation (Asn/Gln) was set as variable modification. For the enriched DTT peptides, oxidation (Met), d0-/d6 DTT (Ser/Thr/Cys), carboamidomethylation (Cys), and deamidation (Asn/Gln) were set as variable modifications. The false discovery rate was set to 0.01 for peptides and proteins. Areas were calculated by Proteome Discoverer 2.0 at 3 ppm mass accuracy. All MS/MS spectra identifying proteins, peptides, or sites were manually inspected for accuracy with reference to Protein Prospector (http://prospector.ucsf.edu/prospector/mshome.htm) according to previously described rules.³⁶

Bioinformatics

Mitochondrial localization annotations of proteins were assigned from at least two prediction software programs including Target P 1.1, Mitoprot, and Predator.^37–39 We also compared our results with a previous data set on the rat heart mitochondrial proteome to confirm its subcellular location. WebLogo⁴⁰ was used to analyze the frequency of amino acids adjacent to the O-GlcNAcylation sites in the peptide sequence. Data sets about the changes of protein abundance between control and diabetic hearts were analyzed by Ingenuity Pathway Analysis (IPA; QIAGEN) for putative upstream regulators and protein interactions.

Immunoblotting of PDH/PDC

The immunoprecipitation of mitochondrial proteins was performed with a commercially available kit (AbCam). Briefly, mitochondrial pellets from control and diabetic hearts were resuspended in 1× PBS buffer supplemented with 1× detergent (AbCam) and 1× Protease Inhibitor Cocktail 1. After being incubated on ice for 30 min, the suspension was spun down at 12 000 × g for 15 min. The supernatant was collected, with protein concentration determined by BCA assay. An equal amount of proteins was incubated with 100 μL of antibody-conjugated resin toward PDH/PDC, with a final volume of 1 mL in PBS. After overnight rotation, the resin was washed five times with 1× Wash Buffer (AbCam). For the CpNag J treatment, immunoprecipitates from control mitochondria were incubated with CpNag J in 20 mM HEPES (pH 7.4) with a protein/CpNag J = 10/1 (w/w) at 37 °C for 2 h. The vehicle buffer (20 mM HEPES (pH 7.4)) was used for samples without CpNag J treatment. The immunocaptured PDH/PDC was released by using the Glycine buffer (AbCam) and then supplemented with 4× Laemmli buffer according to manufacturer instructions. Samples were loaded onto 10% Criterion Gels (Bio-Rad) for SDS-PAGE separation. Proteins in the gels were electro-transferred to a PVDF membrane and probed with RL2 antibody or PDHA1 antibody (AbCam) with the same procedures mentioned above.

■ RESULTS

Functional Assessment of Mitochondria from Control and Diabetic Hearts

Aberrant mitochondrial metabolism plays a fundamental role in the pathogenesis of diabetic cardiomyopathy. Herein we initially determined the functionality of cardiac mitochondria from control and streptozotocin (STZ)-induced diabetic rats. Compared to controls, mitochondria from diabetic hearts show markedly depressed maximum respiratory rates under the oxidative phosphorylation (OXPHOS) status (State 3) by using a 96-well extracellular flux analyzer (Figure 1A,B). Our observation is in line with the previous reports in which substantially decreased respiration was observed in isolated mitochondria from STZ-treated mice by oxygraph chambers⁴¹ and in homogenates of STZ-treated rat hearts.⁴² Moreover, diabetic mitochondria took a longer time to phosphorylate a limited amount of ADP, as evidenced by slower repolarization of ΔΨ_m after ADP addition (Figure 1C,D), suggesting a decreased capacity for ATP synthesis. In addition, maximum ΔΨ_m was lower in mitochondria from diabetic hearts (Figure 1C,D) even though maximal mitochondrial NADH levels were similar in either glutamate/malate or succinate with inhibition of cytochrome oxidase (after the addition of KCN; Figure S1A,B). These results suggest that mitochondrial bioenergetics is significantly impaired in diabetic hearts. Although kinetically impaired, the integrity of isolated mitochondria was largely preserved, as can be judged from the lack of a significant difference in respiratory control ratio (RCR) for mitochondria between groups (Figure 1E). In addition, we observed that mitochondrial Ca²⁺ uptake capacity before the opening of the mitochondrial permeability transition pore (mPTP) was lower in mitochondria from diabetic hearts (Figure 1F), which is consistent with a previous report.⁴³

Figure 1.

Impaired cardiac mitochondrial function in diabetic hearts. Mitochondrial oxygen consumption rates were determined with the presence of glutamate/malate (G/M; A) and succinate (Succ.; B). State 4, without the addition of ADP; State 3, with the addition of ADP. (C, D) Mitochondrial membrane potential was determined with either (C) G/M or (D) Succ. (E) Respiratory control ratios (RCRs) of isolated mitochondria were determined from the rates of State 3 and State 4 oxygen consumption in the presence of G/M. (F) Calculated Ca²⁺ uptake capacity prior to permeability transition pore opening in mitochondria isolated from control and diabetic rats (N ≥ 4; the 2-tailed unpaired Student t test was used to compare different groups, with p < 0.05 being considered significant).

Alterations in Mitochondrial Protein Abundance in Diabetic Heart

By using Percoll-gradient ultracentrifugation, highly purified mitochondria were obtained, for which the purity characterization is shown in Figure S2. To define the global changes in protein abundance in functional cardiac mitochondria between control and diabetic rats, proteomic profiling was performed (as shown in Figure S3). By combining the isotopic Tandem Mass Tag (TMT, Thermo Fisher Scientific) labeling and MS/MS, 352 mitochondrial proteins were identified and quantified (with a false positive discovery rate <0.01; number of corresponding unique peptides ≥2; Table S1). The abundance of 139 proteins (∼40%) was changed by at least 20%. Although 67 proteins (46%) showed increased abundance, 39 proteins (54%) were decreased in abundance. Interestingly, hexokinase, the first rate-limiting enzyme in the glycolysis pathway, and mitochondrial pyruvate carrier protein showed decreased abundance by 50% (Figure 2) in mitochondria from diabetic hearts. Moreover, although the abundance of the pyruvate dehydrogenase complex (PDC) was not changed, pyruvate dehydrogenase kinases were strikingly up-regulated (by >7-fold for PDK4), indicating that mitochondrial glucose oxidation in the diabetic heart might be substantially impaired. However, the abundances of almost all proteins involved in fatty acid transport and oxidation were increased (Figure 2; Table S1), suggesting that diabetic hearts might rely more heavily on fatty acid as their fuel. Indeed, these results support the notion that there is decreased capacity for glycolysis and for pyruvate decarboxylation but concomitant elevation of fatty acid oxidation (FAO) in cardiac mitochondria in diabetes.⁴⁴ Also, the metabolic remodeling from glucose oxidation to fatty acid oxidation in heart has been regarded to be an early event initiating progression of diabetic cardiomyopathy.⁴⁴

Figure 2.

Altered abundance of proteins in several major metabolic pathways in mitochondria between control and diabetic hearts. Proteins with decreased abundance are shown in green, proteins with increased abundance are shown in red, while proteins without apparent changes are shown in blue. HK1, hexokinase-1; MPC1, mitochondrial pyruvate carrier 1; PDH, pyruvate dehydrogenase; PDC, pyruvate dehydrogenase complex; PDK4, pyruvate dehydrogenase kinase 4; CPT1, carnitine O-palmitoyltransferase 1; ACADL, very long-chain specific acyl-CoA dehydrogenase; ECHS, enoyl-CoA hydratase, mitochondrial precursor; HADHA, trifunctional enzyme subunit alpha; HADHB, trifunctional enzyme subunit beta; CS, citrate synthase; ACO2, aconitate hydratase; IDH, isocitrate dehydrogenase; OGDH, 2-oxoglutarate dehydrogenase; SUCLA2, succinyl-CoA ligase; SDHA; succinate dehydrogenase A; SDHB, succinate dehydrogenase B; MDH, malate dehydrogenase; FUM, fumarate hydratase.

In contrast, the abundance of proteins within the OXPHOS system seems to change bidirectionally (Table S1). Notably, 14 proteins in complex I showed strikingly decreased abundance. Considering that complex I is the first site of proton pump coupled to NADH oxidation and transports electrons to complex III, the down-regulation of complex I components could account for the diminished maximal oxygen consumption rate and slower ADP phosphorylation in G/M. Moreover, it could contribute to increased reactive oxygen species production in diabetic cardiac mitochondria. Although 14 proteins in complex IV showed increased abundance, a decreased enzymatic activity was observed in diabetic hearts,²⁶ which might lead to further increased oxidative stress in diabetes. Since our discussion is based mainly on the measured protein level changes, we cannot rule out the possibility that other factors (e.g., post-translational modifications), together with the altered protein abundance, contribute to the defective mitochondrial bioenergetics in diabetic hearts.

IPA analysis was performed to explore the potential upstream regulators for the altered abundance of mitochondrial proteins in control and diabetic hearts. Very interestingly, the signaling network mediated by several transcriptional factors (including peroxisome proliferator-activated receptor α (PPARA), peroxisome proliferator-activated receptor δ (PPARD), and estrogen-related receptor alpha (ESRRA)) are shown to be key contributors to the elevated abundance of components in fatty acid utilization and decreased abundance in other mitochondrial processes (Figure S4). Indeed, in accordance with the previously described regulatory roles of these proteins in diabetic hearts or closely related conditions,^45–47 these results suggest that transcriptional regulation of metabolic pathways plays an important role in mitochondrial dysfunction in chronic diabetic hearts.

It is noteworthy that we did not observe obvious TMT ratio compression in this study, probably due to the relatively less complex organelle proteome (i.e., mitochondrial lysate) and the extensive fractionation of the peptide pool.⁴⁸ Moreover, the gel-free approach we used is especially appealing compared to the gel-based protein profiling for diabetic mitochondria described in previous studies.^49,50 Since mitochondria are rich in basic and hydrophobic proteins, gel-based proteomics is particularly unsuitable for the identification of mitochondrial proteins.⁵¹ Moreover, protein quantification from certain gel spots “identified” as specific proteins may include contributions from other unresolved proteins, and changes in spot density may not reliably reflect actual changes in protein expression since post-translational modifications may alter migration patterns. Indeed, numerous hydrophobic proteins including MPC1, VDAC1/2, and SLC25A4 have been positively identified and quantified in our study. Of note, some of the mitochondrial DNA encoded multipass membrane proteins including MT-ND2 and MT-ND5 have been successfully identified and quantified as well. The increased MT-ND2 and decreased MT-ND5, in combination with the altered abundance of mitochondrial biosynthetic machinery proteins (Table S1), add further evidence to the bidirectional changes of protein abundance contributing to the mitochondrial dysfunction in chronic diabetes.

Extensive O-GlcNAcylation of Mitochondrial Proteins

To gain a preliminary view of mitochondrial protein O-GlcNAcylation, we performed two-dimensional gel separation (i.e., IEF and SDS-PAGE) of the mitochondrial proteins. A number of spots were seen after immunoblotting with CTD 110.6, a commonly used O-GlcNAc-specific antibody (Figure 3A,B). Surprisingly, while the O-GlcNAcylation on some proteins was increased, it was found to be strikingly decreased on others. These observations have been further confirmed by using another pan-specific O-GlcNAc antibody RL2, although a slightly different O-GlcNAc recognition pattern was exhibited (Figure S5).

Figure 3.

O-GlcNAcylation occurs on many mitochondria proteins. Highly purified cardiac mitochondrial proteins from (A) control rats and (B) diabetic rats were separated with two-dimensional gel (IEF/SDS-PAGE), transferred to PVDF membranes, and immunoblotted with CTD 110.6.

Instead of using the 2-D gel-based method for identification of O-GlcNAcylated proteins, we adopted a gel-free approach to determine the molecular identity of the O-GlcNAc proteins in cardiac mitochondria. Specifically, a large-scale O-GlcNAcomic profiling was conducted by combining a refined β-elimination/Michael addition with light/heavy DTT (BEMAD) approach and MS/MS (Figure S3).^33–35 In total, 86 mitochondrial proteins (with 167 sites) were found to be O-GlcNAcylated from control and diabetic hearts (Figure 4), with detailed information on modified peptides shown in Table S2 and Figure S6. Among them, 56 proteins (∼65%) have ≥ 2 O-GlcNAc sites. It is noteworthy that multiple enzymes in major metabolic pathways (including the TCA cycle and fatty acid oxidation) are O-GlcNAc modified. Intriguingly, PDHA1, the α subunit of PDH in PDC, is also O-GlcNAcylated, with two sites mapped.

Figure 4.

O-GlcNAcylated proteins are involved in multiple mitochondrial pathways. O-GlcNAcylated proteins are shown in red, with number of sites shown in parentheses. PDH, pyruvate dehydrogenase; CPT1, carnitine O-palmitoyltransferase 1; CPT2, carnitine O-palmitoyltransferase 2; ACADL, very long-chain specific acyl-CoA dehydrogenase; ECHS, enoyl-CoA hydratase, mitochondrial precursor; HADHA, trifunctional enzyme subunit alpha; HADHB, trifunctional enzyme subunit beta; CS, citrate synthase; ACO2, aconitate hydratase; IDH, isocitrate dehydrogenase; OGDH, 2-oxoglutarate dehydrogenase; SUCLA2, succinyl-CoA ligase; SDHA; succinate dehydrogenase A; SDHB, succinate dehydrogenase B; MDH, malate dehydrogenase; FUM, fumarate hydratase; CI, complex I; CII, complex II; CIII, complex III; CIV, complex IV; CV, complex V; SLC24A4, ADP/ATP translocase 1; SLC25A3, phosphate carrier protein; MCU, mitochondrial Ca²⁺ uniporter; VDAC1/2, voltage-dependent anion-selective channel protein 1/2; SOD2, superoxide dismutase [Mn]; PRDX3, thioredoxin-dependent peroxide reductase; HSP60, 60 kDa heat shock protein; SAM50, sorting and assembly machinery component 50 homologue; CKMT2, creatine kinase S-type, mitochondrial precursor. Please refer to Table S2 for detailed description of the corresponding O-GlcNAc peptides.

In addition, although the sequence analysis of the overall pattern of O-GlcNAcylation sites, as expected, does not show an O-GlcNAcylation consensus motif, the data do indicate a preference for the presence of hydrophobic amino acids prior to the O-GlcNAc site and polar amino acids following it (Figure S7).

Differential O-GlcNAcylation of Mitochondrial Proteins in Control and Diabetic Hearts

The site-specific O-GlcNAc changes in diabetic hearts were determined by the corresponding peak area of peptides containing either light or heavy isotopes. Considering that the abundance of many proteins is markedly changed between control and diabetic hearts, the site-specific O-GlcNAc changes were normalized to the abundance of corresponding proteins (i.e., O-GlcNAc relative site occupancy ratio (ROR)) by using the following equation:

$${\text{ ROR }}_{\text{(Diabetes/Control) }}==\frac{{[\text{O}-\text{GlcNAc}]}_{\text{Diabetes }}/{[\text{ Protein }]}_{\text{Diabetes }}}{{[\text{O}-\text{GlcNAc}]}_{\text{Control }}/{[\text{ Protein }]}_{\text{Control }}}$$

To our surprise, although 45 sites (∼26%) showed increased O-GlcNAcylation by ≥1.2-fold, 68 sites (∼40%) showed decreased O-GlcNAc level by 20% in diabetic mitochondria (Table S3), suggesting dynamic O-GlcNAcylation of mitochondrial proteins. Interestingly, a relatively stable site occupancy (i.e., ROR = 1) was observed for some proteins (e.g., ACAA2 and CKMT2) since the altered O-GlcNAc level (e.g., S357 of ACAA2 and S51 of CKMT2) is proportional to fold-chance of the protein abundance of proteins in diabetic hearts (Table S3). This result suggests that the O-GlcNAc level on certain sites might be directly resulting from the altered abundance of some proteins. Moreover, it seems that the ROR values of a number of other individual O-GlcNAc sites are regulated in a more complicated way independent of protein abundance, as exemplified by S231 of ATP5B, S144 of NDUFB5, and S242 of NNT (Figure 5A). Of note, the bidirectional change also occurs on individual proteins with ≥ 2 sites (e.g., DLAT, CTP1B, HADHA, and HADHB; Figure 5A,B). Even more intriguingly, a number of proteins even showed decreased RORs of all their sites identified, including PDHA1 and SDHA (Figure 5A; Table 3).

Figure 5.

Site-specific O-GlcNAc dynamics on enzymes in (A) pyruvate decarboxylation and TCA cycle and (B) fatty acid utilization in control and diabetic hearts. Only proteins with ≥ 2 O-GlcNAc sites were shown. The dashed line denotes an ROR = 1. PDHA1, pyruvate dehydrogenase subunit alpha; DLAT, dihydrolipoyllysine-residue acetyltransferase component of pyruvate dehydrogenase complex; CPT1B, carnitine O-palmitoyltransferase 1, muscle isoform; HADHA, trifunctional enzyme subunit alpha; HADHB, trifunctional enzyme subunit beta; ACO2, aconitate hydratase; SDHA; succinate dehydrogenase A.

Although overall increases in O-GlcNAcylation in mitochondria from high-glucose-treated cardiomyocytes²⁵ or STZ-induced diabetic hearts²⁶ have been observed previously, the O-GlcNAcomic profiling data presented here suggest that many cardiac mitochondrial proteins are actually differentially O-GlcNAcylated between control and diabetes and that the bidirectional site-specific O-GlcNAcylation changes are partially the result of the protein abundance changes. Mislocalization of the O-GlcNAc transferase within cardiac mitochondria in diabetic hearts could be another contributor to this phenomenon.²⁶ Since the cross-talk between O-GlcNAcylation and phosphorylation has been detected on multiple nuclear and cytoplasmic proteins,^52,53 changes in mitochondrial phosphoproteomic pattern might also affect the site-specific O-GlcNAc levels within mitochondria. The results of an investigation into this point will provide deeper insight into the bidirectional O-GlcNAc changes in diabetic mitochondrial proteins (Junfeng Ma et al., manuscript in preparation). The unanticipated complexity of the regulation on the occupancy level of specific O-GlcNAc sites implies that differential O-GlcNAcylation status may site-specifically alter the functions of individual mitochondrial proteins.

Validation of O-GlcNAcylation of Pyruvate Dehydrogenase (PDH)

PDHA1, the α subunit of PDH, is the major known regulatory factor that governs the conversion of pyruvate to Acetyl-CoA.⁵⁴ PDH was immunoprecipitated from rat cardiac mitochondria lysate and probed with RL2 for PDHA1. Indeed, PDHA1 is O-GlcNAcylated even in control hearts (Figure 6). Moreover, ∼20% decreased O-GlcNAcylation was observed in diabetic cardiac mitochondria (Figure 6). To further confirm the O-GlcNAc status of PDHA1, immunoprecipitated PDH was treated by CpNag J (Clostridium perfringens Nag J), a recombinant homologue of human O-GlcNAcase.⁵⁵ As expected, CpNag J treatment potently removed O-GlcNAc from PDHA1 (Figure 6). Taken together, these results suggest that PDHA1 is O-GlcNAcylated, and its O-GlcNAcylation might be directly involved in the regulation of PDH and thus the metabolic dysregulation in the development of diabetic cardiomyopathy.

Figure 6.

Validation of O-GlcNAcylation of PDHA1 in cardiac mitochondria from control and diabetic rats. Pyruvate dehydrogenase complexes were immunopurified from lysates of cardiac mitochondria from control and diabetic rats. After extensive washing, the immunoprecipitates were treated with or without CpNag J as indicated, then separated with SDS-PAGE, and transferred to PVDF membranes for blotting with PDHA1 and RL2. Blots in the presence of 1 M GlcNAc were used as a competitive assay.

■ DISCUSSION

Studies on nuclear and cytoplasmic O-GlcNAcylation have suggested that aberrant protein O-GlcNAcylation is an important contributor to the diverse adverse effects on the diabetic heart. The recent identification of mitochondrial O-GlcNAc cycling components (i.e., OGT, O-GlcNAcase, and a UDP-GlcNAc transporter), the altered enzymatic activity of mOGT and O-GlcNAcase, and the mislocalization of OGT suggest that mitochondrial O-GlcNAcylation can be an underlying factor in the progression of diabetic cardiomyopathy.²⁶

In the present study, we systematically investigated the bioenergetics of mitochondria isolated from control and STZ-induced diabetic hearts. In comparison to controls, the diabetic cardiac mitochondria show diminished oxygen consumption rates, slower ADP phosphorylation rates, and impaired recovery of ΔΨ_m (especially in steady-state respiration status), suggesting inefficient ATP production. Providing that sustained heart function is completely dependent on mitochondrially derived ATP to fuel contraction, the reduced metabolic efficiency should have detrimental effects on cardiac function. Moreover, the lower Ca²⁺ activation threshold for mPTP opening might further exacerbate the progressive decline of cardiac function in diabetes.

The cardiac proteomes of the functionally active mitochondria were stable isotope-labeled and analyzed by advanced MS methods, yielding a pattern of protein abundance in diabetes that is different from controls. The results from our comprehensive profiling of the mitochondrial proteome provide extensive evidence that there is metabolic remodeling in diabetic hearts, from glucose utilization to fatty acid oxidation, and this precedes the onset of cardiac dysfunction. Moreover, the bidirectional changes in the abundances of other proteins (including OXPHOS proteins) afford detailed insights into the impaired bioenergetics and accompanying mitochondrial dysfunction in diabetic hearts.

Mitochondrial O-GlcNAcomics, in control and diabetic hearts, was enabled through the use of a refined BEMAD approach. We find that many cardiac mitochondrial proteins are O-GlcNAcylated. The O-GlcNAcylated proteins are involved in multiple mitochondrial pathways including pyruvate decarboxylation, fatty acid oxidation (FAO), TCA cycle, oxidative phosphorylation, mitochondrial protein import machinery, and others. Since many of these processes are proposed to be correlated with insulin resistance, we hypothesize that O-GlcNAcylation of mitochondrial proteins is likely related to insulin resistance in chronic diabetes. Moreover, unexpectedly, O-GlcNAc site occupancy on mitochondrial proteins is subject to bidirectional changes (i.e., either up-regulated or down-regulated) between the control and diabetic hearts, and O-GlcNAcylation at certain sites of individual proteins can even be reduced in diabetic hearts. This seemingly paradoxical observation casts doubt on the notion that protein O-GlcNAcylation could be uniformly elevated with the induction of hyperglycemia. Several possible mechanisms might explain this finding: (1) O-GlcNAc transferase is abnormally localized in the mitochondria from diabetic hearts, perhaps causing its inability to O-GlcNAcylate certain sites on proteins; (2) the aberrant mitochondrial O-GlcNAc cycling in chronic diabetes might substantially affect the O-GlcNAcylation of its targets; (3) O-GlcNAcylation modifies many kinases, and the modification might activate certain mitochondrial kinases, thus decreasing O-GlcNAcylation due to competition;^3,52,53 (4) akin to protein phosphorylation, different O-GlcNAc sites of individual proteins may respond differentially to certain physiological or pathological stimuli, and this site-specific O-GlcNAcylation might consign specific roles to proteins in contributing to mitochondrial (dys)function. Therefore, it is very likely that the dynamic O-GlcNAc cycling of individual mitochondrial proteins is more complicated than previously expected.

By using immunoprecipitation and Western blotting with pan-specific antibody, we validated the O-GlcNAc status of PDHA1, a key subunit of the enzyme PDH which regulates the conversion of pyruvate to Acetyl-CoA. Indeed, PDHA1 is O-GlcNAcylated in control hearts, which was then confirmed by the treatment with CpNag J, a recombinant homologue of human O-GlcNAcase. The potent removal of O-GlcNAc on PDHA1 by CpNag J also suggests that mitochondrial proteins could be potential targets for O-GlcNAcase.

In comparison to control hearts, the overall O-GlcNAcylation level of PDHA1 was ∼20% decreased in diabetic hearts. This change could be due to the crosstalk between O-GlcNAcylation and phosphorylation since not only the overall increased phosphorylation,⁵⁴ but also site-specifically increased phosphorylation on PDH have been observed in diabetic hearts (Junfeng Ma et al., manuscript in preparation). Considering that the increased phosphorylation, together with other factors,⁵⁶ decreases the activity of PDH/PDC and thus diminishes glucose utilization, we reason that O-GlcNAcylation should serve a regulatory role in cardiac metabolism, contributing to the progression of diabetic cardiomyopathy. The determination of the O-GlcNAc stoichiometry of PDHA1 is underway for detailed delineation of its functional importance.

As the most abundant post-translational modification, glycosylation has been found on numerous proteins.⁵⁷ Although a few mitochondrial proteins were regarded to be N-glycosylated,^58–60 whether there is a bona fide N-glycosylation synthetic machinery is still not widely accepted, and the characterization of the corresponding glycosyltransferases and hydrolases is incomplete. Herein, we show that O-GlcNAcylation exists on many mitochondrial proteins, suggesting that O-GlcNAcylation is probably the major glycosylated form on proteins within mitochondria. Moreover, the O-GlcNAcyation mediated by the O-GlcNAc cycling components (all of which are present in mitochondria)²⁶ is a dynamic process, exerting important regulatory roles in mitochondria.