Removal of Abnormal Myofilament O-GlcNAcylation Restores Ca²⁺ Sensitivity in Diabetic Cardiac Muscle

In diabetic cardiomyopathy, the contractile and electrophysiological properties of the cardiac muscle are altered (1). Prior studies have mainly focused on alterations in Ca²⁺ handling (2–4). However, these perturbations alone unlikely account for lower force production and altered relaxation typically found in the heart of patients with diabetes (5,6). Indeed, the intrinsic properties of cardiac myofilaments appear to be altered too (7,8). More specifically, Ca²⁺ sensitivity (ECa²⁺₅₀), a measure of myofilament force production at near physiological Ca²⁺ levels, is reduced in the heart of patients with diabetes (9–11). Yet the mechanisms responsible for Ca²⁺ desensitization in diabetic hearts remain incompletely understood.

O-GlcNAcylation is a posttranslational modification (PTM) linked to glucose metabolism and centrally involved in regulating cellular homeostasis (12). This PTM consists of the addition of single O-linked β-N-acetyl-d-glucosamine (O-GlcNAc) sugar to serine (Ser) and threonine (Thr) residues of nuclear and cytoplasmic proteins. The reaction is catalyzed by O-GlcNAc transferase (OGT), whereas O-GlcNAc removal is under the control of O-GlcNAcase (OGA) (12). Excessive O-GlcNAcylation results from glucose- or other nutrient-induced overload of the hexosamine biosynthesis pathway (HBP). Alterations in HBP are increasingly recognized as a major contributing factor for insulin resistance (12) and “glucose toxicity” during diabetes. Similar to phosphorylation, O-GlcNAcylation is a widely distributed and highly dynamic PTM (12). However, unlike phosphorylation, which is regulated by a myriad of kinases and phosphatases, the extent of O-GlcNAcylation relies on two enzymes only, specifically OGT and OGA. OGT substrate specificity is regulated by transient protein-to-protein interactions that take place primarily at its tetratricopeptide repeat (TPR) domain (12) (Fig. 7A). Often OGT and OGA interact with each other and/or are found forming a holoenzyme complex with protein phosphatases and kinases (12). Modifications of Ser and Thr by O-GlcNAc occur in myofilaments, and the addition of exogenous N-acetyl-d-glucosamine (GlcNAc) alters myofilament response to Ca²⁺ (13,14). In addition, manipulation of cardiac O-GlcNAc levels influences Ca²⁺ cycling kinetics (4) and mitochondrial rates of respiration in diabetes (2,15) and functional recovery after ischemia-reperfusion injury (16,17) or chronic pressure-overload (18–21). Despite all of this evidence, whether and how altered O-GlcNAcylation contributes to myofilament dysfunction in diabetic cardiomyopathy (7,8,22–24) is currently unclear.

Here we proved that specific removal of O-GlcNAc excess from diabetic myofilaments ameliorates contractile dysfunction by linking improvement in force-Ca²⁺ relationships to site-specific O-GlcNAc changes. We also determined a potential mechanism leading to altered O-GlcNAcylation by comparing the status and sarcomeric distribution patterns of OGT and OGA in the heart of rats with streptozotocin (STZ)-induced type 1 diabetes with that found in controls.

All animal protocols in this study were performed in accordance with institutional guidelines and approval of the Johns Hopkins University School of Medicine Institutional Animal Care and Use Committee.

Type 1 diabetes was induced in male Sprague-Dawley rats (Charles River) by an intraperitoneal injection of STZ (65 mg/kg). Control animals were injected with vehicle only. Animals were killed 6–8 weeks after induction of diabetes. At the moment of tissue harvest, STZ diabetic animals had blood glucose of 693.5 ± 61.5 mg/dL and controls had 123.9 ± 9.5 mg/dL.

For skinned cardiac muscles studies, muscles were isolated and mounted as previously described (25,26). Varied Ca²⁺ concentrations [Ca²⁺]_o were achieved by mixing the relaxing solution and activating solution in various ratios. After reaching the highest [Ca²⁺]_o concentration, trabeculae were washed in relaxing solution and incubated at room temperature 1 h in 1 µg/mL CpOGA (27) diluted in relaxing solution. Afterward, a new Ca²⁺ activation protocol was repeated. Steady-state force-[Ca²⁺]_o relationships were determined experimentally and fit to a modified Hill equation (25,26).

Cold acetone–fixed cryosections (7–8 μm) from rat or human (BioChain Inc., Office of Human Research Protection registered IRB00008283) myocardium were blocked and incubated overnight with a primary antibody against O-GlcNAc (CTD110.6). In addition, isolated skinned myocytes were obtained from flash-frozen myocardium by homogenization in 0.03% Triton X-100 at low speed as previously described (28), seeded on eight chamber slides coated with 40 μg/mL laminin (Invitrogen), and fixed in 4% formaldehyde-methanol–free ultrapure (Polysciences Inc.). Cryosections or isolated skinned myocytes were blocked and incubated overnight with OGA O-GlcNAcase (345), OGT (AL-25) (29), and anti–α-actinin (Sigma-Aldrich) at 1 μg/mL. Secondary antibodies were Alexa 647 goat anti-mouse IgM (μ-chain) for O-GlcNAc, Alexa 647 goat anti-rabbit IgG for OGT, Alexa 594 goat anti-chicken IgY for OGA, and Alexa 488 goat anti-mouse IgG for α-actinin. Prolong antifade with DAPI (Invitrogen) was used for mounting. Images were acquired on a Zeiss 710 LSM upright microscope using a ×25 or a ×63 water-immersion objective (Nikon) and analyzed with Zen 9 Leica Zeiss software tools.

For gold immunolabeling, goat anti-rabbit or goat anti-mouse were labeled with 12-nm-diameter particles to detect anti-OGT (AL-34) or anti–O-GlcNAc (CTD 110.6), and goat anti-chicken was labeled with 6-nm-diameter particles to detect anti-OGA (345). Labeled ultrathin sections (60- to 90-nm thick) were examined under the transmission electron microscope (Hitachi 7600 TEM, Tokyo, Japan), and 8–11 random field pictures were used for quantification of OGT, OGA, and O-GlcNAc immunolabeled gold particles with ImageJ software (National Institutes of Health [NIH]).

For immunoprecipitation studies, anti-OGT (AL-28) or anti-OGA (345) (1 μg total) antibodies were added to 0.5 mg/mL protein samples. Immunoprecipitates were then separated by SDS-PAGE, transferred to polyvinylidene fluoride membranes (Millipore, Bedford, MA) for antibody probing against α-cardiac actin (Sigma-Aldrich), α-tropomyosin (Tm; Sigma-Aldrich), and anti-myosin light chain (MLC) 1 (Clone 1LC-14, Spectral Diagnostics). Between different antibodies blots were stripped for 1 h at 25°C in 200 mmol/L glycine (pH 2.5; Sigma-Aldrich) (29).

Heart homogenates were separated in cytosolic or myofilament fractions, desalted, and subjected to OGT activity assays as described (30). Activity counts in disintegration per minute (dpm) were normalized to total protein content (dpm/μg), and then background activity without CKII peptide was subtracted (Fig. 7C).

OGAse activity was determined as previously described (31). Briefly, activity was expressed as the amount of enzyme catalyzing the release of 1 mmol/μg/min pNP from pNP-GlcNAc, and then background activity in the presence of the most specific and potent OGA inhibitor (1μmol/L thiamet-G) was subtracted (Fig. 7C).

Myofilament proteins were isolated as previously described (32). Protein concentration was determined by the Bradford assay and equal amounts of protein (∼200 μg) were reduced with 5 mmol/L dithiothreitol (DTT), alkylated with 15 mmol/L iodoacetamide, and digested by trypsin (trypsin-to-protein ratio, 1:50). Tryptic peptides were labeled for quantitation with tandem mass tags (TMT) 10plex labels (127, 128, 129, 130) following the manufacturer’s guidelines (Thermo Fisher Scientific). The labeled peptides were combined and fractioned offline using XBridge HPLC column (Waters). The resulting 96 fractions were combined into 24 fractions for liquid chromatography–tandem mass spectrometry (LC-MS/MS) runs, while only reserving 10% for the preenriched analysis, and combining the remaining 90% for O-GlcNAc peptide enrichment (postenrichment).

The remaining pooled peptides were used to enrich O-GlcNAcylated peptides by adapting a method described previously (33,34), with some modifications. In brief, peptides were treated with alkaline phosphatase (50 units; New England Biolabs) and PNGase F (1,000 units; New England Biolabs) for 6 h, followed by desalting with a C18 spin column (Nest Group). The dried peptides were resuspended in a buffer containing 20 mmol/L DTT, 20% (v/v) ethyl alcohol, and 1.5% triethanolamine (pH 12.5) and incubated at 50°C for 4 h with gentle shaking. Reaction was quenched by the addition of trifluoroacetic acid (final pH ∼7.0). Peptides were desalted and then incubated with thiol-Sepharose beads (Sigma-Aldrich) in PBS containing 1 mmol/L EDTA (PBS/EDTA, pH 7.4) for 4 h. After five washes in a PBS/EDTA buffer supplemented with acetonitrile 40% (v/v), beads were incubated in PBS/EDTA containing 20 mmol/L DTT for 30 min. Released peptides were collected and desalted with a C18 spin column. Dried peptides were then analyzed by LC-MS/MS for O-GlcNAc site mapping and quantification.

The unenriched and O-GlcNAc–enriched fractions were analyzed with an LTQ-Orbitrap Velos (Thermo Fisher Scientific) attached to a Nano Acquity (Waters) chromatography system. Peptides were loaded on a 75-μm × 2.5-cm C18 (YMC*GEL ODS-A 12-nm S–10-μm) trap at 600 nL/min 0.1% formic acid (solvent A) and fractionated at 300 nL/min on a 75-μm × 150-mm reverse-phase column using a 2–90% acetonitrile in 0.1% formic acid gradient over 90 min. Eluting peptides were sprayed into the mass spectrometer through 1-µm emitter tip (New Objective) at 2.2 kV. Survey scans (full MS) were acquired within 350–1800 m/z with up to eight peptide masses (precursor ions) individually isolated at IW1.9Da, and fragmented (MS/MS) using higher-energy collisional dissociation, at 35 activation collision energy. Precursor and the fragment ions were analyzed at resolution 30,000 and 15,000, respectively. Dynamic exclusion of 30 s, repeat count 1, monoisotopic ion precursor selection (MIPS) “on”, m/z option “off”, lock mass “on” (silocsane 371 Da) were used. MS/MS spectra were processed by PEAKS Studio (Bioinformatics Solutions Inc.) using Rattus norvegicus FASTA as the proteome database, with concatenated decoy database, specifying all peptide species, trypsin as enzyme, missed cleavage 2, precursor mass tolerance 10 ppm, fragment mass tolerance 0.03 Da, and oxidation (M), deamidation (NQ), carbamidomethyl (C), and TMT labels 127, 128, 129, and 130 as variable modifications.

Quantitation function of PEAKS Studio was used to export the raw intensity values of TMT peptides with or without O-GlcNAc enrichment. Only peptides with a positive identification value or a cutoff value of 20 for peptide score threshold (−10 log P), quantification mass tolerance (0.2 Da), and a 0.1% false discovery rate (FDR) were considered. All Ser or Thr residues from O-GlcNAcylated peptides were identified by a +136 mass shift given by DTT during BEMAD and their MS/MS spectra manually inspected. The experimental design consisted of 12 biological samples divided into three experiments with four isobaric mass-tags, each experiment comprising a STZ diabetic and a control sample with and without CpOGA treatment. The TMT preenrichment samples were analyzed to factor in potential changes in protein expression, the relative total protein load, and to refine the comparisons among postenrichment samples by accounting for technical and experimental variation. The median signal value for each modified peptide of each sample was first determined and converted to log2 notation (with 0.0 values excluded as nulls) for further processing. Data were then quantile normalized to achieve the same median. The difference between prenormalized and normalized data provided a number that was further used as a correction factor for the postenrichment peptide analysis. For the O-GlcNAc–enriched peptides, signals were treated as above, except when converted to log2 notation, the 0.0 values were log-converted to 0.001 to be able to express ratios.

We used a statistical generalized linear model approach for differential gene expression detection. Briefly, a mixed-effects linear model was fit for each individual modified peptide to estimate O-GlcNAcylation differences between the groups of samples being compared (i.e., control vs. STZ diabetic, control vs. CpOGA control, STZ vs. CpOGA STZ). When distinct peptides for the same O-GlcNAcylated site were available, correlation coefficients were computed and the associated consensus correlation was added to the model (35). An empirical Bayes approach was applied to moderate standard errors of log2 O-GlcNAcylation fold-change, as previously described (36). Finally, for each analyzed feature, moderated t statistics, log-odds ratios of differential expression (B statistics), and raw and adjusted P values (FDR control by the Benjamini and Hochberg method) were obtained. All analyses were performed using software packages available from the R/Bioconductor for statistical computing “limma” (36).

The Student t test, one-way ANOVA, and two-way ANOVA with repeated measures, followed by post hoc pairwise comparison, when appropriate, was used for statistical analysis of the data. A value of P < 0.05 was considered to indicate significant differences between groups. Unless otherwise indicated, data are expressed as mean ± SEM.

Using an MS approach, we have previously identified specific O-GlcNAcylation sites on five major cardiac myofilament proteins on normal hearts (14). We also showed that incubation of skinned cardiac muscles with GlcNAc reduces myofilament Ca²⁺ sensitivity, thus reproducing a hallmark of diabetic cardiac muscle (7,14). Yet, site-specific O-GlcNAcylation changes on diabetic hearts remain unknown. Nor is it clear whether removing endogenous O-GlcNAc from diabetic skinned cardiac muscles is sufficient to restore myofilament function. To address these questions, we first confirmed that O-GlcNAcylation is enhanced in protein extracts from STZ diabetic hearts (Fig. 1A). Next, we used CpOGA to remove O-GlcNAc from diabetic skinned cardiac muscles (Fig. 1B). Finally, we used a refined global quantitative proteomic technique that combines TMT labeling with β-elimination/Michael addition to quantify the O-GlcNAcylation changes in diabetic hearts and to identify the key amino acid residues where O-GlcNAc is reduced after CpOGA in STZ diabetic hearts but not in controls (Fig. 1C).

OGA is the mammalian enzyme that removes O-GlcNAc in vivo; however, its glycosidase activity is greatly reduced when expressed as a recombinant protein (31). To circumvent this issue, we used a bacterial glycosidase, CpOGA (27) (Clostridium perfringens N-acetyl-glycosidase J, provided by Dr. Daan van Aalen, University of Dundee). CpOGA is highly homologous to human OGA and displays potent and specific activity toward mammalian protein homogenates (37). Hence, skinned cardiac muscles from diabetic (STZ-induced) and control rats were incubated with CpOGA and their contractile properties determined before and after this treatment. Steady-state force-[Ca²⁺] relationships measurements revealed that CpOGA removal of abnormal O-GlcNAcylation restores myofilament Ca²⁺ sensitivity in diabetic skinned fibers (Fig. 1B). Remarkably, the ECa²⁺₅₀ of CpOGA-treated diabetic skinned fibers became similar to that found in control muscles, regardless of CpOGA presence (Fig. 1B). Upon CpOGA administration, no difference in maximal Ca²⁺ activated force (F_max) and Hill coefficient (n) was evident between control and diabetic muscles (Fig. 1B). Next, control and diabetic (STZ) rat heart myofilaments were compared before and after being incubated with recombinant CpOGA and analyzed for O-GlcNAcylation by MS/MS tagging and LC-MS/MS. In total we found 63 O-GlcNAcylated sites, 39 in myosin heavy chain (MHC), 9 in α-sarcomeric actin, 2 in MLC 1, 5 in α-Tm 1, 7 in cardiac troponin I (cTnI), and 1 in myosin binding protein C (Table 1).

Because CpOGA removes O-GlcNAc in both control and diabetic (STZ) rat muscles but only restores ECa²⁺₅₀ in diabetic muscles, we focused our proteomic analysis on identifying the site-specific O-GlcNAc changes that are statistically significant on CpOGA-treated STZ diabetic versus CpOGA-untreated STZ diabetic myofilaments but not significant on CpOGA-treated control versus CpOGA-untreated control myofilaments (Fig. 1C). Surprisingly, most of the sites that change significantly are located on MHC (S740, S844, S1414, S1465, S1471, S1472, S1598, T1601, S1602, S1778, and S1917), α-sarcomeric actin (actin S54, T326), and α-TmS87. Noteworthy, the sites that are significantly more O-GlcNAcylated in diabetic hearts (i.e., MHC S844, S1471, S1472, T1601, and S1917 and actin T326) (Fig. 1C, red rectangles) are also part of the group of sites that change significantly upon CpOGA treatment of STZ diabetic myofilaments. Thus, abnormal O-GlcNAcylation of cardiac muscle proteins is sufficient to reduce myofilament Ca²⁺ sensitivity, and its removal is necessary to prevent this Ca²⁺ desensitization.

The α-cardiac actin O-GlcNAc signal intensity (by immunoblot) is augmented in animal models of diabetes (14). Moreover, in normal human myocardium, O-GlcNAc predominantly modifies a Z-line protein called ZASP (Z-band alternatively spliced PDZ motif protein) (38). ZASP O-GlcNAcylation increases further in heart failure and hypertrophic cardiomyopathy (38). However, whether O-GlcNAc–specific subcellular localization changes in diabetes is not clear. To fill this gap, cryosections from normal and STZ-induced diabetic rat hearts were fixed in cold acetone and stained for O-GlcNAc (CTD 110.6) to determine the relative abundance and localization of O-GlcNAc (Fig. 2A). Consistent with previous findings, O-GlcNAc signal intensity was markedly increased in STZ rat diabetic myocardium compared with control hearts (Fig. 2B). Also, immunoelectron microscopy revealed that in control myocardium, O-GlcNAc was predominantly localized in clusters near the Z-lines, whereas in STZ-treated rats O-GlcNAc clusters were spread toward the A-band (Fig. 2C). Next, we performed a morphometric analysis of nine random fields of control or STZ rat diabetic myocardium using immune-gold particles (12 nm diameter) to assess O-GlcNAc distance from the nearest Z-line (Fig. 2D). This approach allowed us to conclude that overall O-GlcNAc signal is increased in the sarcomeres of diabetic rats and moves away from Z-lines. To test whether similar alterations pertain to human pathology, O-GlcNAc immune-fluorescence was investigated in one sample from a normal (donor) and one from a patient with type 2 diabetes. Consistent with findings in the rat, O-GlcNAc signal intensity was markedly increased in the human diabetic specimen (Fig. 2E).

Next, we reasoned that the presence of enhanced O-GlcNAcylation in specific sarcomere compartments of diabetic hearts could be due to enhanced activity of OGT, reduced OGA activity, or abnormal localization of enzymes. To address this issue, we used anti-OGT (AL-28) or anti-OGA (345) antibodies and an immunofluorescence and confocal microscopy approach to analyze the signal intensity and colocalization of both enzymes. OGT exhibited a predominantly sarcomeric localization, whereas OGA displayed both a sarcomeric and reticular pattern (Fig. 3A and B). Similar to O-GlcNAc, OGT tended to colocalize mostly with α-actinin near the Z-line (Fig. 3C and D). Although detectable at the Z-line too, OGA was instead predominantly distributed throughout the A-band along the entire sarcomeric unit. To further consolidate this evidence, we used the colocalization tools of Zen 9 Image Analysis Software (Leica Zeiss), enabling us to quantify changes in OGT and OGA distribution. Representative images of normal and STZ diabetic hearts were used to generate XY pixel dot plots of laser scanning signals. A series of colocalization quantification parameters are displayed as bar graphs for OGT and α-actinin (Fig. 3E) and for OGT and OGA (Fig. 3F). Analysis of at least three random fields from four replicates on three different hearts showed clear differences between normal and STZ diabetic myocardium. OGT colocalization with α-actinin was decreased in STZ myocardium (Fig. 3E1–3), thus confirming a redistribution apart from the Z-line (P = 0.0053). Interestingly, OGT/OGA colocalization was also reduced in STZ diabetic myocardium (P ≤ 0.05, Fig. 3F1–3). α-Actinin, OGT, and OGA immunofluorescence was also evaluated in human heart samples (Fig. 4A). Acetone prefixed frozen slides were triple-antigen stained by immunofluorescence for OGT (AL-28), OGA (345), and α-actinin. As expected, human control and diabetic myocardium exhibited a differential sarcomeric pattern for OGT and OGA signals. A selected area from control (Fig. 4A, top panel) or diabetic (Fig. 4A, bottom panel) is shown as a white square in the merged left picture, followed by enlarged areas showing the regions in single or combined channels for linear surface profile plots of approximately six to eight sarcomere units (dashed line). Combined signals of OGT (magenta) or OGA (red) and α-actinin (green) for control and diabetic myocardium indicated that OGT labeling peaks in control myocardium correspond mostly with the α-actinin signal. Conversely, OGA labeling showed variable signals, peaking both at A-band and Z-line regions (Fig. 4A, top panel). In diabetic myocardium, however, OGT signal peaks tended to shift away from the α-actinin signal, whereas OGA labeling centered more on the α-actinin signal (Fig. 4A, bottom panel). Because cryosections fixed with acetone might reflect tissue architecture alterations, we decided to corroborate the OGT and OGA staining pattern on isolated skinned myocytes obtained from flash-frozen myocardium derived from rat samples as described (28). Similar to rat or human myocardium cryosections, isolated skinned myocytes fixed with 4% formaldehyde displayed the expected OGT and OGA staining pattern in control (Fig. 4B, top panel) or STZ diabetic (Fig. 4B, bottom panel) myocytes. When sarcomeres in control hearts are compared with those found in diabetic hearts, these changes in the OGT staining pattern mirror faithfully those found in rat STZ diabetic myocardium (Fig. 3), with an inverted pattern of OGT and OGA sarcomeric distribution. This inverted pattern of OGT/OGA localization may influence O-GlcNAcylation cycling rates, thus likely perturbing O-GlcNAc stoichiometry in specific sarcomere compartments.

Next, we analyzed OGT and OGA by double immunoelectron microscopy in hearts from STZ diabetic and control rats. Immunoelectron microscopy data fully corroborated the evidence obtained with the immunofluorescence approach, confirming the redistribution of both enzymes within the sarcomere. Indeed, similar to O-GlcNAc, OGT immunogold-labeled particles were mainly located at the Z-disk in normal hearts (Fig. 5A, top panel), whereas they were more diffused along the A-band in diabetic hearts (Fig. 5B, top panel). Instead of being localized mainly at the A-band (normal hearts), OGA formed sizable clusters in the vicinity of the Z-disk (diabetic hearts; Fig. 5B, bottom panel). Finally, we determined the frequency of OGT and OGA immunogold-labeled particles, examining 9–10 random fields of normal or diabetic myocardium and quantifying OGT (12 nm, purple circle) and OGA (6 nm, red circle) immunogold-labeled particles confined to the myofilament apparatus. Our approach revealed that both O-GlcNAc cycling enzymes were detected at higher frequency in diabetic myocardium (Fig. 5C and D). Taken together, these data suggest that, in analogy to cardiac kinases and phosphatases (39), mislocalized OGT and OGA activities can affect the function of contractile or regulatory proteins, or both.

Assessing OGT by Western blot in myofilament preparations (14) involves the use of high (1%) concentrations of Triton X-100 (32). This procedure may preclude the possibility of detecting weak interactions between OGT, OGA, and their potential transient binding partner proteins, thus underestimating the extent of OGT and OGA abundance in the myofilaments. To properly evaluate these interactions, we used a coimmunoprecipitation approach, testing whether OGT and OGA are physically associated with myofilament proteins and determining possible changes in their myofilament abundance imparted by diabetes. Fresh whole-heart homogenates from control or diabetic (STZ-treated) rats were immunoprecipitated with anti-OGT (AL-28) or anti-OGA (345) antibodies (29), resolved by SDS-PAGE, and analyzed by Western blots against several myofilament proteins. OGT and OGA were both associated with α-cardiac actin, α-Tm, and MLC 1 in normal hearts (Fig. 5E and F). A representative coimmunoprecipitation of OGT and OGA with interacting myofilament proteins is provided in Fig. 5E and F. In the diabetic hearts, the OGA-immunoprecipitated interactions with α-cardiac actin, α-Tm, and MLC 1 were increased several fold compared with the control interactions (Fig. 5H) (n = 5 vs. n = 4, P < 0.05), whereas interactions with OGT were normal. On the other hand, although the OGT-immunoprecipitated interactions with myofilaments tended to have increased interactions in diabetic samples, they did not reach statistical significance. The heterogeneity of the signals for OGT and OGA coimmunoprecipitations on the control heart homogenates might be related to the feeding status (ad libitum) of the animals or to a potential cyclic nature of these protein-to-protein interactions. Hence, OGA interactions with myofilament proteins are more abundant in diabetic hearts, and they are likely contributing to reduce contractility.

In diabetic hearts, overall O-GlcNAc levels have been reported to change with or without reciprocal changes in OGT expression and uridine diphosphate (UDP)–GlcNAc levels (40). Protein specific O-GlcNAcylation, however, can change in either direction (40). With this paradox in mind, we hypothesized that the observed changes in OGT/OGA sarcomeric distribution and myofilament protein interactions might influence their enzymatic activity. To test this, fresh whole hearts from control or STZ diabetic rats were homogenized in extraction buffer containing 1% Triton X-100. Total tissue homogenates were separated in cytosolic and myofilament fractions by centrifugation. For OGT assays, both cytosolic and myofilament fractions were desalted into OGT desalting buffer (for activity assays) or stored at −80°C for SDS-PAGE and Western blots. After OGT expression was normalized to cTnI content, neither subcellular fraction showed significant differences between control (n = 3) and STZ diabetic (n = 3) hearts (myofilament and cytosolic fractions are shown in Fig. 6A and B, respectively). OGT activity was assayed by the incorporation of UDP–[H³]-labeled GlcNAc into a synthetic peptide (CKII peptide); disintegration per minute counts were normalized to micrograms of protein, and then OGT activity without peptide was subtracted (31) (Fig. 7C). This approach revealed a significant reduction in myofilament-associated OGT activity in STZ diabetic hearts (Fig. 6D, obtained in myofilament fractions). Conversely, OGT activity in the cytosolic fractions of STZ diabetic hearts was not significantly different (Fig. 6E).

For OGA assays, myofilament fractions retained traces of detergent; therefore, they were not suitable for measurement of enzymatic activity. However, total tissue homogenates became properly usable after protein precipitation with 30–50% ammonium sulfate and desalting into OGA assay buffer. After the latter procedure, OGA expression normalized to cTnI content was not significantly different between groups (0.54 ± 0.02 for STZ vs. 1.14 ± 0.29 for controls, n = 3 each) (Fig. 6C). OGA activity was assessed by the release of p-nitrophenol (pNP) from pNP-GlcNAc, a synthetic substrate. In addition, our approach accounted for OGA activity background more strictly because we subtracted OGA activity detected in the presence of thiamet-G, the most potent and specific OGA inhibitor available (41) (Fig. 7C). OGA-specific activity was significantly increased in homogenates from STZ diabetic hearts compared with controls (Fig. 6F). Thus, changes in OGT/OGA protein-to-protein interactions and subcellular localization may influence OGT/OGA enzymatic activity.

This study establishes that excessive O-GlcNAcylation, in a specific subset of myofilament sites of MHC, α-sarcomeric actin, and Tm, is sufficient to produce myofilament dysfunction in diabetic cardiomyopathy, and it does so by affecting Ca²⁺ sensitivity. These perturbations in myofilament O-GlcNAcylation result from OGT and OGA displacement within the sarcomere rather than from variations in enzymatic activity per se, resulting in an overall perturbed O-GlcNAc cycling. Important, from a translational point of view, is the finding that similar alterations occur in experimental and human diabetic hearts.

Depressed myofilament Ca²⁺ sensitivity is a hallmark of myofilament dysfunction in diabetic cardiomyopathy and heart failure (7,11). Alterations in myofilament phosphorylation are now recognized as important negative modulators of cardiac function, along with perturbations in Ca²⁺ handling (39,42–44). That alterations in cardiac protein O-GlcNAcylation are present in models of type 1 and type 2 diabetes (2,4,24,40,45) and cardiac disease (21,38) is well established; however, no studies have addressed yet the functional effect of cardiac myofilaments O-GlcNAcylation during the course of diabetic cardiac dysfunction (2–4). We have previously identified numerous sites within the cardiac myofilament are O-GlcNAcylated under normal conditions (14). We also showed that incubation of normal skinned cardiac muscles with GlcNAc reduced myofilament Ca²⁺ sensitivity (14). Moreover, adenoviral-based or inducible transgenic overexpression of human OGA, for which CpOGA is a very close homolog, is known to ameliorate contractile and energetic deficits associated with diabetic cardiomyopathy (4,45). Here, we add important novel evidence that in diabetic myofilaments there are selective proteins, such as MHC, α-sarcomeric cardiac actin (actin), and cardiac α-Tm, and some of their specific sites such as MHC S844, S1471, S1472, T1601 and S1917, actin T326, and TmS87, that indeed have excessive O-GlcNAcylation. More importantly, we demonstrated that specific removal of excessive O-GlcNAcylation, from a subset of myofilaments sites via CpOGA (Fig. 1C) rapidly restores myofilament Ca²⁺ sensitivity, thus correcting myofilament dysfunction in STZ diabetic skinned muscles. The advantage of our approach (i.e., the use of recombinant CpOGA on skinned cardiac muscles) resides in the ability of assessing the functional consequences of O-GlcNAcylation removal from myofilaments independently from other major adverse effects eventually imposed by O-GlcNAcylation on Ca²⁺ handling and mitochondrial key proteins (2,4,15,24,45–48). Thus, here we cement the view that excessive O-GlcNAcylation is a novel, negative modulator of myofilament function in the diabetic myocardium. At the same time, our data suggest, for the first time, that removing excess of O-GlcNAc from diabetic myofilaments specific sites mimics what dephosphorylation does on Ca²⁺ sensitivity in experimental (49) and human heart failure (50–53).

Cardiac myofilament phosphorylation and function is regulated by multiple kinases and phosphatases (54,55). They strategically dock at the Z-line and modify their activity and/or localization in response to mechanical forces and neurohumoral stimuli (54,55). The displacement of these enzymatic activities may have important functional repercussions on the contractile apparatus (39). For instance, at baseline, protein phosphatase 2A forms a complex with protein phosphatase 2B and p38–mitogen-activated protein kinase, but it moves away from the Z-line upon β-adrenergic stimulation (55). Whether similar effects can be attained by changes in the sarcomeric localization of OGT and OGA is currently unknown.

Here we report that, in normal hearts, OGT and OGA distribute mainly at the Z-line and A-band, respectively. However, in diabetic hearts, OGT drifts away from the Z-line, whereas OGA appears even more clustered at this site. Importantly, these changes are similarly evident in rat and human diabetic myocardium. It is already well known that OGT localization and catalytic activity changes in response to metabolic cues (12,56). OGT and OGA often occur in transient protein-to-protein complexes containing kinases and phosphatases (12,56). These include kinases and phosphatases such as AMPK (56), Ca²⁺/calmodulin-dependent protein kinase IV, p38–mitogen-activated protein kinase, protein phosphatase 1, and myosin phosphatase targeting 1, which are key regulators of cardiac metabolism and function (12). Although these protein interactions have not been confirmed in the heart, a recent report elegantly shows that CaMKII is O-GlcNAcylated during acute hyperglycemia and diabetes, leading to chronic activation, which contributes to diabetes cardiac mechanical dysfunction and arrhythmias (24). Present data show that, in the heart, OGT and OGA form complexes with α-cardiac actin, α-Tm, and MLC 1, and that these protein-to-protein complexes are enhanced for OGA in diabetes. These alterations are potentially relevant to the pathogenesis of diabetic systolic and diastolic dysfunction, given that transient protein-to-protein complexes regulate OGT substrate specificity (12). Because O-GlcNAcylation and phosphorylation signaling cascades have extensive cross-talk (12,56), both at the level of site occupancy and at the level of modifying enzymes, further dissecting this phenomenon in the heart would be important.

Previously, OGT expression (40,57,58) and activity (59) in diabetic hearts have been found to be normal, whereas OGA activity is reduced, with or without reciprocal changes in expression (57–59). Here, we analyzed myofilament and cytosolic fractions separately. We found that OGT expression in diabetic hearts is not different between diabetic and control hearts. Owing to OGA inhibition by traces of Triton X-100, OGA expression and activity were analyzed only in total homogenates in which OGA expression was not different between groups. In addition, our study accounts for OGT and OGA activity background in a more rigorous way. Indeed, OGT activity without CKII peptide and OGA activity with thiamet-G (a specific OGAse inhibitor) were subtracted. Our myofilament versus cytosolic compared data showed that OGT enzymatic activity is reduced in the myofilament in diabetic hearts but not in the cytosolic fraction. In contrast, OGA activity in total homogenates is significantly increased in diabetic hearts (Fig. 6F). This apparent paradoxical decrease of OGT activity in diabetic hearts could be explained by the way we accounted for the background activity, subcellular fractionation, or by changes in OGT substrate specificity in response to metabolic cues (30). Increased OGA activity may reflect compensation for excessive O-GlcNAcylation. Supporting this possibility is previous evidence showing that increased erythrocyte protein O-GlcNAcylation and OGA activity are present in samples from patients with prediabetes and with type 2 diabetes (60).

Cardiac contractility is regulated on a beat-to-beat basis by the Ca²⁺-dependent modulation of myosin cross-bridge binding on actin by the Tm-troponin complex. Cross-bridge cycling occurs at the A-band (61), whereas most of the signaling arises from the Z-line (54,55,62). Thus, retargeting of OGT and OGA may increase α-cardiac actin O-GlcNAcylation at the A-band, potentially interfering with actin-Tm modulation of cross-bridge cycling (Fig. 7B). Altogether, these data suggest that in diabetic hearts, OGT and OGA displacement rather than their catalytic activity is the key in modulating excessive O-GlcNAcylation.

There are limitations in the current study. First, more work is needed to define the functional effect of myofilament O-GlcNAcylation at the site-specific level, along with the interplay with phosphorylation. Second, future studies should address whether excessive α-cardiac actin O-GlcNAcylation indeed perturbs cross-bridge cycling kinetics, actin-Tm interactions, or the actin rate of polymerization. Finally, it is not known whether O-GlcNAc modifications or phosphorylation of specific substrates or in OGT/OGA contribute to their altered pattern of localization. Answering this question would define more in detail the effect of this PTM on the function of diabetic hearts; however, it requires fully dedicated, separate studies.

In conclusion, the present work demonstrates that abnormal O-GlcNAcylation is sufficient to cause myofilament functional deficit in diabetic cardiomyopathy. Increased OGA interactions with sarcomeric proteins (α-cardiac actin, α-Tm, and MLC 1) are likely central to these alterations, showing that normal sarcomeric OGT and OGA subcellular localization is lost in myofilaments from diabetic hearts. This abnormal redistribution of O-GlcNAc cycling enzymes is similarly present in experimental and human diabetic cardiomyopathy. On these grounds, here we propose that, in addition to direct targeting of abnormal site-specific phosphorylation of myofilament regulatory subunits, removing abnormal site-specific O-GlcNAcylation in myofilaments and/or preventing changes in O-GlcNAc cycling should be considered as another promising new therapeutic avenue for treating diabetic cardiomyopathy.

Acknowledgments. For expert technical assistance, the authors thank John Robinson from Pediatric Cardiology and the JHMI Microscope Facility, and Marina Allary (Johns Hopkins School of Medicine) for helpful suggestions.

Funding. This work has been supported by the National Institutes of Health (NIH)/National Heart, Lung, and Blood Institute (NHLBI) Proteomics Center Contract HHSN268201000032C (A.M.M., G.W.H.), R01-DK-61671 (G.W.H.); Program of Excellence in Glycobiology NHLBI P01-HL-107153 (G.W.H., N.P.), R01-HL-091923 (N.P.); American Heart Association (AHA) and the Lawrence and Florence A. DeGeorge Charitable Trust Scientist Developing Grant (AHA-12SDG9140008) for G.A.R.-C., AHA (AHA-0855439E) for W.D.G.; and Institutional Development Award (IDeA) from National Institute of General Medical Sciences/NIH (P20-GM-12345 and R01-DK100595) for C.S. In addition, this study was supported by the Johns Hopkins Institute for Clinical and Translational Research, which is funded in part by Grant Number UL1-TR-001079 from National Center for Advancing Translational Sciences/NIH for L.M.

Duality of Interest. Under a licensing agreement between Johns Hopkins University and several companies, including Covance Research Products, Sigma-Aldrich, and Santa Cruz Biotechnology, G.W.H. receives royalties from the sale of the CTD110.6 O-GlcNAc antibody. The terms are managed by Johns Hopkins University in accordance with conflict of interest policies. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. G.A.R.-C. designed and performed research, analyzed data, and wrote the manuscript. J.M. and C.S. designed and performed research, analyzed data, and edited and reviewed the manuscript. Q.Z. performed research, analyzed data, and edited and reviewed the manuscript. N.S.L.-F., M.X., V.C., K.C., L.D., and R.N.C. performed research and analyzed data. X.S. performed research. L.M., N.P., and G.W.H. assisted with experimental discussion and with critical evaluation and editing of the manuscript. W.D.G. analyzed data and edited and reviewed the manuscript. A.M.M. designed research, interpreted data, and assisted with experimental discussion and with critical evaluation and editing of the manuscript. G.A.R.-C. is the guarantor of this work and, as such, had full access to all data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Prior Presentation. Parts of this study were presented at the Basic Cardiovascular Sciences 2011 Scientific Sessions of the American Heart Association, New Orleans, LA, 18–21 July 2011, and at the American Heart Association 2012 Scientific Sessions, Los Angeles, CA, 3–7 November 2012.