A reference dataset of O-GlcNAc proteins in quadriceps skeletal muscle from mice

Abstract

A key nutrient sensing process in all animal tissues is the dynamic attachment of O-linked N-acetylglucosamine (O-GlcNAc). Determining the targets and roles of O-GlcNAc glycoproteins has the potential to reveal insights into healthy and diseased metabolic states. In cell studies, thousands of proteins are known to be O-GlcNAcylated, but reference datasets for most tissue types in animals are lacking. Here, we apply a chemoenzymatic labeling study to compile a high coverage dataset of quadriceps skeletal muscle O-GlcNAc glycoproteins from mice. Our dataset contains over 550 proteins, and > 80% of the dataset matched known O-GlcNAc proteins. This dataset was further annotated via bioinformatics, revealing the distribution, protein interactions, and gene ontology (GO) functions of these skeletal muscle proteins. We compared these quadriceps glycoproteins with a high-coverage O-GlcNAc enrichment profile from mouse hearts and describe the key overlap and differences between these tissue types. Quadriceps muscles can be used for biopsies, so we envision this dataset to have potential biomedical relevance in detecting aberrant glycoproteins in metabolic diseases and physiological studies. This new knowledge adds to the growing collection of tissues with high-coverage O-GlcNAc profiles, which we anticipate will further the systems biology of O-GlcNAc mechanisms, functions, and roles in disease.

Introduction

Tissues must dynamically respond to nutrient levels in their environment to maintain metabolic homeostasis (O'Brien 2022). A mechanism conserved in all animal cells is dynamic glycosylation of intracellular proteins with O-linked N-acetylglucosamine (O-GlcNAc) (Wells et al. 2003; Hardivillé and Hart 2014). O-GlcNAcylation is a regulatory protein post-translational modification (PTM) that enables tissue to rapidly sense nutrient levels within minutes of glucose flux (Marshall et al. 2004). Knowledge of O-GlcNAc is extensive in cellular models, with nearly 8,000 human and 5,000 mouse proteins known to carry this modification discovered across more than three decades of in vitro studies (Hart 2014; Ma et al. 2021; Wulff-Fuentes et al. 2021; Ma et al. 2022). A key knowledge gap remains in mapping O-GlcNAc proteins in human and animal tissues. High coverage O-GlcNAc studies have only been performed on brain, (Trinidad et al. 2012; Wang et al. 2017; Burt et al. 2021) heart, (Narayanan et al. 2023) and placenta (Luna et al. 2024) as of this writing. This report presents a reference set of O-GlcNAc glycoproteins enriched from mouse skeletal muscle as a step toward filling this knowledge gap and enabling studies of O-GlcNAc roles in muscle physiology and disease.

Some of the challenges of O-GlcNAcylation studies are the dynamic nature of this modification, its low stoichiometry, and its tendency to fragment in mass spectrometry (MS) ionization conditions used for proteomics (Ma and Hart 2017). In all metazoan cells, just two proteins regulate the addition and the removal of O-GlcNAc sugars, the O-GlcNAc transferase (OGT) and the O-GlcNAcase (OGA) (Fig. 1a). OGT and OGA can cycle O-GlcNAc modifications onto and off proteins on the minutes timescale (Song et al. 2008; Yang et al. 2008; Dias et al. 2009). OGT and OGA are each promiscuous enough to use thousands of cellular proteins as substrates. Perhaps because of this promiscuity, efficiencies of O-GlcNAcylation for specific proteins vary between substrates, and the stoichiometry of O-GlcNAc modified vs. unmodified proteins varies. It is estimated that 3%–12% of the expressed copies of the average OGT substrate protein is O-GlcNAcylated in standard conditions, (Darabedian et al. 2018; Leturcq et al. 2018) though some proteins like nucleoporin 62 (Nup62) are observed to be nearly 100% glycosylated (Rexach et al. 2010). Interestingly, tissue differences in O-GlcNAc stoichiometry have been observed. For example, cyclic AMP response element binding protein (CREB) is measured to have 45% of expressed protein as the O-GlcNAcylated species in rat cerebellum, compared with 33% of CREB being O-GlcNAcylated in rat liver (Rexach et al. 2010). Because of the wide range of stoichiometries of O-GlcNAc across the proteome, it is important to pre-enrich O-GlcNAcylated proteins using antibodies, chemical labels, or enzymatic labels prior to proteomic analysis in order to obtain a high coverage dataset (Fehl and Hanover 2022; Nelson et al. 2023).

Fig. 1.

Mapping O-GlcNAc glycoproteome from mouse quadriceps muscle. a) O-GlcNAc protein modifications are dynamically cycled by OGT and OGA in response to glucose and other nutrients. b) State of the art of rodent O-GlcNAc proteins in skeletal and cardiac muscle. c) Workflow for O-GlcNAc profiling used in this study. Monosaccharide symbols follow the SNFG (symbol nomenclature for Glycans) system. (Varki et al. 2015).

Skeletal muscle is an extensive metabolic organ, consuming a significant proportion of total glucose pools in all vertebrates, including mice and humans (Merz and Thurmond 2020). Importantly, skeletal muscle insulin resistance is a precursor to diabetes and can predate diabetes by up to 10 years (Greene et al. 2018). Because glucose use, insulin signaling, and O-GlcNAc are highly interconnected, we hypothesized that determining a high coverage muscle O-GlcNAc dataset would be useful for both physiology and metabolic disease studies. Some knowledge of O-GlcNAc in skeletal muscle has already been obtained from rat studies, where Cienewkski-Bernard et al found 14 O-GlcNAc proteins in rat gastrocnemius muscle, (Cieniewski-Bernard et al. 2004) Hédou et al found 4 additional O-GlcNAc proteins from rat soleus muscle, (Hédou et al. 2009) and Ma et al found 7 more O-GlcNAc proteins found from rat hearts (Ma et al. 2016) for a total of 25 O-GlcNAc muscle proteins. In muscle cell lines, however, Deracinois et al found 342 O-GlcNAc proteins in mouse C2C12 myotube model cells (Deracinois et al. 2018). Because of the high number of proteins in muscle cell line determined to carry O-GlcNAc, we expected to find a higher number of proteins in skeletal muscle tissue than previously reported. To our best knowledge, no publication has reported a high-coverage O-GlcNAc proteome in mouse skeletal muscle such as the quadriceps.

In this study, we isolated 559 O-GlcNAcylated proteins from mouse quadriceps skeletal muscle, closing the knowledge gap in O-GlcNAcome of muscle cells vs. muscle tissue (Fig. 1b). We used a chemoenzymatic strategy to label and enrich O-GlcNAc glycoproteins, a system optimized by the Hsieh-Wilson lab (Khidekel et al. 2003; Thompson et al. 2018). After extensive bioinformatic analysis to confirm that our 559 enriched proteins had >80% overlap with known O-GlcNAc glycoproteins, we also compared our quadriceps dataset to a recently reported high-coverage O-GlcNAc map of mouse hearts (Narayanan et al. 2023). We expect that this reference of O-GlcNAc proteins in skeletal muscle will add to the growing collection of O-GlcNAc tissue databases to enable improved systems biology studies (Ma et al. 2022).

Results

Preparation of freshly extracted mouse quadriceps muscle.

We used wild type C57BL/6J mice for this reference dataset because this strain has been used to model human muscle physiology (Jacobs et al. 2013). Mice were raised on a standard low-fat diet and sacrificed after 6 months. Quadriceps muscle was removed, homogenized, and proteins were extracted for an initial assessment of O-GlcNAc glycosylation patterns. We used an extraction buffer containing detergent for the extraction of diverse proteins, as well as the potent OGA inhibitor thiamet-G (Yuzwa et al. 2008) to inhibit rapid deglycosylation of muscle proteins from released OGA.

To enrich O-GlcNAcylated proteins from muscle tissue, we applied a chemoenzymatic labeling strategy (Fig. 1c) (Khidekel et al. 2003). Briefly, the glycosyltransferase enzyme β-1,4-galactosyltransferase (GalT) was used to selectively label protein O-GlcNAc sites. The GalT Y289L mutant can use an azide-labeled uridine diphospho-N-acetylgalactosamine (UDP-GalNAz) derivative as a substate, installing GalNAz as a chemical label onto a diverse range of protein GlcNAc sites. (Torres and Hart 1984) GalT(Y289L) expressed well in bacteria and was active for labeling proteins following reported methods (Supp. Fig. 1a) (Thompson et al. 2018). With azide-labeled O-GlcNAc proteins from muscle samples in hand, we applied copper-catalyzed azide-alkyne cycloaddition (CuAAC) “click” chemistry to install biotin enrichment handles. Streptavidin-horse radish peroxidase (Strep-HRP) was also used to visualize O-GlcNAcylated proteins (Supp. Fig. 1b) as we as positive and negative labeling controls (Supp Fig. 1c). The bands on these western blots revealed hundreds of proteins ranging from 17 kDa – 250 kDa, consistent with the ca. 5000 O-GlcNAcylated proteins found in the “O-GlcNAcome” database (Wulff-Fuentes et al. 2021). Because most of the O-GlcNAcome comes from in vitro cellular studies, we set out to match the quadricep tissue profile to known O-GlcNAcylated proteins using mass spectrometry-based proteomics.

Proteomic profiling of the native O-GlcNAc profile of quadriceps muscle soluble proteins

Proteins labeled with biotin tags are readily enriched using streptavidin magnetic beads. (Liu et al. 2022) For protein identification workflows, quadriceps tissue was homogenized and soluble proteins were extracted and labeled with UDP-GalNAz. We next used CuAAC to attach biotin-alkyne, then enriched biotin-labeled glycoproteins on streptavidin beads. Following glycoprotein enrichment, we conducted on-bead reduction, alkylation, and digestion with the endoproteases Lys-C and trypsin. Peptides were separated with nano-liquid chromatography (nano-LC) and detected with tandem mass spectrometry (MS/MS). All peptides were also analyzed with phosphorylation as a variable modification so as not to miss phosphopeptides, a common feature on signaling proteins (Vlastaridis et al. 2017).

Our goal was to observe the widest array of O-GlcNAc glycoproteins in these skeletal muscle tissue samples, so we classified hits if a protein had at 2 or more unique peptides or at least 1 phosphopeptide identified in at least 1 mouse. The total number of proteins identified was 559 (Fig. 2a). The full dataset is compiled in Appendix 1, with sheet 1 containing all 559 hits, and Sheet 2 containing the 63 phosphoproteins. An overlap with the O-GlcNAcome (Wulff-Fuentes et al. 2021) and the O-GlcNAcAtlas (Ma et al. 2021) datasets showed 82% of this dataset were known O-GlcNAc proteins (Fig. 2b).

Fig. 2.

Analysis of mouse quadriceps O-GlcNAc protein interactions and cellular location. a) Comparison of the 559 proteins identified in our study with two databanks of O-GlcNAc proteins in Mus musculus, the O-GlcNAcome (accessed 2024 may 12) and the O-GlcNAcATLAS (version 3). b) Comparison with combined O-GlcNAc proteomics studies from gastrocnemius (Cieniewski-Bernard et al. 2004), soleus (Hédou et al. 2009), and cardiac (Ma et al. 2016) muscles. Note: These prior studies were all performed on rat tissue. c) Subcellular compartment analysis of the 559 proteins. Bars indicate the mean with 95% confidence interval. unique and shared protein IDs are listed, with enriched proteins from shared proteins exceeding a fold change of +/− 0.5 and p value <0.05 cutoff. n = 3. See Appendix 1 for the full dataset.

O-GlcNAc and phosphorylation PTMs are known to have cross-talk on the same protein, potentially mediating alternative activity and signaling features (Wang et al. 2010; Leney et al. 2017; Bourré et al. 2018; Chen et al. 2018). Among the identified hits, 63 proteins had phosphopeptides observed (Appendix 1, Sheet 2) and the total number of identified phosphorylation sites in these 63 proteins is 106 (i.e. some O-GlcNAc proteins have multiple phosphorylation sites identified). All 63 phosphoproteins carry both O-GlcNAc and phosphorylation concurrently, offers an opportunity to study the interplay of O-GlcNAc with phosphorylation. Metabolic enzymes comprised the primary group of phospho/O-GlcNAc proteins, including glyceraldehyde 3-phosphate dehydrogenase (GAPHD), triosephosphate isomerase 1 (TPI1), aldolase A (ALDOA), phosphoglucose mutase (PGM), aldehyde dehydrogenase 2 (ALDH2), and ATP synthase subunit 5B (ATP5B). Another major group included muscle proteins like myosin-binding protein C (MYBPC2) and troponin T (TNNT3), raising the possibility that O-GlcNAc and phosphorylation regulate skeletal muscle functions and metabolism.

If we applied more conservative hit-calling parameters, in which each protein had at least 3 unique peptides and was identified in at least 2 mice, we obtained 222 “high-confidence” hits (Supp. Fig. 2) (dataset in Appendix 1, Sheet 3). Among the subset of more conservative hits, we observed 79% overlap with known O-GlcNAc proteins (Supp. Fig. 2c). This ratio was similar to the 82% overlap in the main dataset of 559 hits.

Notably, we detected several constitutively (ca. 100% glycosylated) O-GlcNAcylated proteins including the nucleoporins Nup62, Nup153, and Nup214, which were viewed as positive controls. (Comer et al. 2001) We also observed tau, (Yuzwa et al. 2008) host cell factor 1 (HCFC1), (Lazarus et al. 2013) and myosin phosphatase target subunit 1 (MYPT1)(Pedowitz et al. 2021) as other well-known O-GlcNAcylated proteins as further confidence in our labeling experiment. In all mice, we observed GAPDH, a key metabolic housekeeping protein involved in glycolysis with known O-GlcNAc regulatory sites. (Park et al. 2009).

Within the 559 hits, 101 proteins did not overlap with known O-GlcNAc proteins in either the O-GlcNAcATLAS (Ma et al. 2021) or the O-GlcNAcome databank (Wulff-Fuentes et al. 2021). These 101 proteins could be either false positives or potentially unreported O-GlcNAcylated proteins. Gene ontology (GO) enrichment analysis of these 101 proteins was dispersed across many gene families, and did not have strong GO term enrichment, so we plan to save the analysis of these 101 proteins for more extensive scrutiny, which will involve a larger number of replicates analysis in future studies.

We also compared the proteins identified in this dataset from quadriceps muscle with previously known O-GlcNAc glycoproteins in other types of skeletal muscle tissue. Combined, the publicly available datasets of prior skeletal muscle tissue O-GlcNAc studies included 25 proteins, all from rat skeletal muscles (Cieniewski-Bernard et al. 2004; Hédou et al. 2009; Ma et al. 2016). Our dataset overlapped with 16 of these proteins (70%) (Fig. 2b). These prior studies were conducted in the gastrocnemius, (Cieniewski-Bernard et al. 2004) soleus, (Hédou et al. 2009) and cardiac muscles (Ma et al. 2016) of rats, which may explain the 7 non-overlapping hits between these former studies and the present study in mouse quadriceps muscle. Our dataset with 559 skeletal muscle O-GlcNAc glycoproteins presents an over 20-fold increase in the number of O-GlcNAc glycoproteins in skeletal muscle tissue from any species.

To validate the proteomics results, we performed immunoblotting on a second set of quadriceps extracts from each mouse as an independent verification. We repeated the O-GlcNAc chemoenzymatic labeling, but following enrichment on biotin beads we eluted directly into gel electrophoresis running buffer. The eluates from each mouse were immunoblotted with antibodies to probe for proteins previously reported to have O-GlcNAc modifications. For known O-GlcNAc proteins, we confirmed strong staining even of understudied O-GlcNAc proteins like oxoglutarate dehydrogenase (OGDH) and fatty acid binding protein 3 (FABP3). To investigate some of the 101 proteins that we isolated that have not been previously reported, we immunoblotted for three of these: calmodulin (CALM3), glycogen synthase 1 (GYS1), and Rad23 homolog B (Rad23B). Each of these proteins revealed a strong band in the O-GlcNAc enrichment (Fig. 2c). We noted the Rad23B antibody cross-reacted with the Rad23A protein, which is a known O-GlcNAc protein. Though a full validation of the remaining 101 proteins remains beyond the scope of this work, a high-throughput technique such as electron-dissociation transfer mass spectrometry (ETD-MS) may facilitate faster validation of the 101 potentially novel O-GlcNAc proteins.

Bioinformatic analysis of muscle O-GlcNAc proteins

One of the limitations of O-GlcNAc analysis using chemoenzymatic biotin tagging is an increased difficulty in assigning the modification site following enrichment because the glycopeptides can remain bound to the resin. We therefore used alternative informatic methods to ascertain whether the isolated proteins were O-GlcNAc glycoproteins in addition to overlap with previously reported O-GlcNAc proteins (Fig. 2a).

A protein’s physical location in a cell can be used to assess O-GlcNAc on soluble nucleocytoplasmic and mitochondrial proteins vs. other types of glycoproteins like N-glycoproteins. Here, we rely on the fact O-GlcNAc proteins are primarily intracellular and localized in the nucleus, cytoplasm, and mitochondria, (Wulff-Fuentes et al. 2021) unlike canonical N-glycans and O-glycan oligosaccharide modifications found on membrane-associated proteins (Baycin Hizal et al. 2014). Analysis of subcellular localization is an established method to bioinformatically validate O-GlcNAc glycoprotein enrichment relative to other glycans that could have been labeled by the GalT enzyme, as has been reported (Luna et al. 2024). When we analyzed the locations of our hits using the STRING database, (Szklarczyk et al. 2019) we observed the distribution of these proteins between cellular compartments (Fig. 2d). The top three subcellular locations were the cytosol, nucleus, and mitochondria, as expected for O-GlcNAc glycoproteins.

We also analyzed the protein–protein interactions (PPI) of these 559 hits using the STRING database, because we have observed selective O-GlcNAc modification of individual proteins in multiprotein complexes (Liu et al. 2022). The PPI network was constructed using experimentally verified interactions, and is extensive due to the large number of proteins identified (Supp Fig. 3). We observed a large set of 271 linked interacting proteins, 25 proteins that each made only one or two interactions, and 258 proteins that were singletons (non-interactors). The proteins that made only one interaction were typically related isoforms, like glycerol-3-phopshate dehydrogenase isoforms 1 and 2 (GPD1 and GPD2) and lactate dehydrogenase A and B (LDHA and LDHB) that may share other protein interactions or act as oligomeric complexes (Damavandi et al. 2023).

Fig. 3.

Analysis of mouse quadriceps O-GlcNAc protein biological functions. a) Significant gene ontology (GO) biological process terms for the 559 proteins. Circle size is proportional to the number of associated proteins (full list shown in Supp. Fig. 5). GO analysis and network were performed with the ClueGO application in Cytoscape. b) Percentage breakdown of GO biological processes in the mouse quadriceps muscle O-GlcNAc proteins. Analysis performed in the ClueGO app in Cytoscape.

The largest multi-protein interactomes that we discovered included protein complexes involved in chaperone/folding activity, glycolysis, muscle function, fatty acid metabolism, and tubulin structure (Supp. Fig. 4). The chaperone cluster included heat shock proteins like heat shock protein 90 (HSP90) and HspB, disulfide isomerase proteins like protein disulfide isomerase A3 (PDIA3), and the DnaJ proteins. O-GlcNAcylation has been shown to regulate a variety of chaperone proteins including heat shock and DnaJ proteins (Javed et al. 2024). The glycolysis cluster included GAPDH, enolase 3 (ENO3), TPI1, and phosphatidylethanolamine-binding protein 1 (PEBP1), each of which are known O-GlcNAc proteins in the O-GlcNAc database (Wulff-Fuentes et al. 2021). Fatty acid metabolic proteins included fatty acid binding protein 3 (FABP3), hydroxy-CoA dehydrogenase (HADH), and acetyl-CoA acetyltransferase 2 (ACAA2), which may mediate fatty acid metabolism in response to hexosamine nutrient levels. Muscle proteins included myosin chains, titin, and vimentin, many of which also interacted with a subcluster of tubulin-associated proteins like tau and the 14-3-3 proteins Ywhae, Ywhaz, and Ywhag. These interaction clusters revealed the potential for O-GlcNAc to regulate key muscle processes via modulating protein–protein interactions.

Fig. 4.

Comparison of reported whole heart O-GlcNAc study with our quadriceps data. a) Scheme for the method used by the Zachara lab (Narayanan et al. 2023). b) overlap of quadriceps proteins (this study) with heart proteins. 82% of the quadriceps proteins and 34% of the heart proteins are annotated as O-GlcNAc proteins (Narayanan et al. 2023). c) top 10 gene ontology terms and p-values for the heart and quadriceps O-GlcNAc enrichment studies.

Gene ontology analysis of functional O-GlcNAcome of quadriceps muscle

We explored the functional roles of the proteins identified using a gene ontology (GO) enrichment analysis with the ClueGO extension in Cytoscape (Bindea et al. 2009). In ClueGO, we used the setting corresponding to the “100% detail” (compared to “100% global”) in order to focus on specific GO processes. We applied the statistical enrichment cutoff of P = 0.05 to discover 43 GO processes, the top 20 of which are shown in Table 1 and the full list shown in Supplementary Table 1. The top three GO terms were glucose metabolism, aerobic electron transport chain, and canonical glycolysis. These top three GO terms are each key energetic processes known to be dynamically regulated by O-GlcNAcylation (Gonzalez-Rellan et al. 2022). Glutamine catabolism was the #5 GO process, consistent with O-GlcNAc regulatory roles in the hexosamine biosynthetic pathway, which incorporates glutamine as the nitrogen donor for hexosamine synthesis (Wellen et al. 2010). Regulation of long-chain fatty acid metabolism was the #8 GO term, with FABP3 being a key protein involved that is known to be O-GlcNAcylated in response to high fat conditions (Pritt et al. 2008; Zhang et al. 2013). The #7 and #9 GO terms were muscle-specific GO processes, actomyosin contraction and skeletal muscle contraction via calcium regulation, respectively. Our dataset is consistent with published GO analyses of the overall O-GlcNAcome, which show similar enriched terms including the regulation of sugar and amino acid metabolism (Hart 2019). Unique GO features to our dataset additionally included regulation of skeletal muscle contraction, nerve/muscle action potential, and calcium signaling (Fig. 3a). These processes are critical in skeletal muscle, which, due to its large tissue size, is involved in regulating whole-body energy balance.

Table 1.

Top gene ontology enrichment terms for the 559 quadriceps O-GlcNAcome. Full list shown in Supplementary Table 1.

ID	Term	Number of Genes	Term PValue	-log10 (p-value)	Associated Genes Found
GO:0006007	glucose catabolic process	10.00	9.663E-12	11.015	[Actn3, Aldoa, Eno3, Gapdh, Ldha, Pfkm, Pgam2, Pgk1, Pkm, Tpi1]
GO:0061621	canonical glycolysis	8.00	1.253E-10	9.902	[Aldoa, Eno3, Gapdh, Pfkm, Pgam2, Pgk1, Pkm, Tpi1]
GO:0019646	aerobic electron transport chain	6.00	6.049E-08	7.218	[Coq9, Dld, Ndufa10, Ndufs3, Park7, Uqcrc1]
GO:0042775	mitochondrial ATP synthesis coupled electron transport	6.00	1.762E-06	5.754	[Coq9, Dld, Ndufa10, Ndufs3, Park7, Uqcrc1]
GO:0009065	glutamine family amino acid catabolic process	3.00	4.595E-04	3.338	[Fah, Got1, Got2]
GO:0043650	dicarboxylic acid biosynthetic process	3.00	8.941E-04	3.049	[Got1, Got2, Mthfd1]
GO:0000915	actomyosin contractile ring assembly	2.00	1.335E-03	2.875	[Pdcd6ip, Plec]
GO:0140212	regulation of long-chain fatty acid import into cell	2.00	1.335E-03	2.875	[Eprs, Fabp3]
GO:0014809	regulation of skeletal muscle contraction by regulation of release of sequestered calcium ion	2.00	1.335E-03	2.875	[Casq1, Dmd]
GO:0019551	glutamate catabolic process to 2-oxoglutarate	2.00	1.335E-03	2.875	[Got1, Got2]
GO:0019550	glutamate catabolic process to aspartate	2.00	1.335E-03	2.875	[Got1, Got2]
GO:0072249	metanephric podocyte development	2.00	3.907E-03	2.408	[Lamb2, Lamc1]
GO:1901526	positive regulation of mitophagy	2.00	3.907E-03	2.408	[Slc25a4, Slc25a5]
GO:0005471	ATP:ADP antiporter activity	2.00	3.907E-03	2.408	[Slc25a4, Slc25a5]
GO:1901841	regulation of high voltage-gated calcium channel activity	2.00	3.907E-03	2.408	[Dysf, Nipsnap2]
GO:0035308	negative regulation of protein dephosphorylation	2.00	7.624E-03	2.118	[Cmya5, Ywhae]
GO:0032222	regulation of synaptic transmission, cholinergic	2.00	7.624E-03	2.118	[Camk2b, Lama2]
GO:0140021	mitochondrial ADP transmembrane transport	2.00	7.624E-03	2.118	[Slc25a4, Slc25a5]
GO:1902915	negative regulation of protein polyubiquitination	2.00	7.624E-03	2.118	[Dysf, Plaa]
GO:0019661	glucose catabolic process to lactate via pyruvate	2.00	7.624E-03	2.118	[Actn3, Ldha]
GO:0006114	glycerol biosynthetic process	2.00	7.624E-03	2.118	[Got1, Pgp]
GO:0003069	acetylcholine-mediated vasodilation involved in regulation of systemic arterial blood pressure	1.00	3.658E-02	1.437	[Sod2]
GO:0035402	histone kinase activity (H3-T11 specific)	1.00	3.658E-02	1.437	[Chek1]
GO:0042247	establishment of planar polarity of follicular epithelium	1.00	3.658E-02	1.437	[Wdr1]
GO:0051232	meiotic spindle elongation	1.00	3.658E-02	1.437	[Ppp2r1a]
GO:0006122	mitochondrial electron transport, ubiquinol to cytochrome c	1.00	3.658E-02	1.437	[Uqcrc1]
GO:0003117	regulation of vasoconstriction by circulating norepinephrine	1.00	3.658E-02	1.437	[Snta1]
GO:1900826	negative regulation of membrane depolarization during cardiac muscle cell action potential	1.00	3.658E-02	1.437	[Cav3]
GO:2000469	negative regulation of peroxidase activity	1.00	3.658E-02	1.437	[Gstp1]
GO:0009298	GDP-mannose biosynthetic process	1.00	3.658E-02	1.437	[Mpi]
GO:0002037	negative regulation of L-glutamate import across plasma membrane	1.00	3.658E-02	1.437	[Arl6ip5]
GO:1903946	negative regulation of ventricular cardiac muscle cell action potential	1.00	3.658E-02	1.437	[Bin1]
GO:1904878	negative regulation of calcium ion transmembrane transport via high voltage-gated calcium channel	1.00	3.658E-02	1.437	[Bin1]
GO:0032804	negative regulation of low-density lipoprotein particle receptor catabolic process	1.00	3.658E-02	1.437	[Anxa2]
GO:1905602	positive regulation of receptor-mediated endocytosis involved in cholesterol transport	1.00	3.658E-02	1.437	[Anxa2]
GO:0006175	dATP biosynthetic process	1.00	3.658E-02	1.437	[Adk]
GO:0016005	phospholipase A2 activator activity	1.00	3.658E-02	1.437	[Plaa]
GO:1903384	negative regulation of hydrogen peroxide-induced neuron intrinsic apoptotic signaling pathway	1.00	3.658E-02	1.437	[Park7]
GO:0046295	glycolate biosynthetic process	1.00	3.658E-02	1.437	[Park7]
GO:0036482	neuron intrinsic apoptotic signaling pathway in response to hydrogen peroxide	1.00	3.658E-02	1.437	[Park7]
GO:0031448	positive regulation of fast-twitch skeletal muscle fiber contraction	1.00	3.658E-02	1.437	[Actn3]
GO:0009257	10-formyltetrahydrofolate biosynthetic process	1.00	3.658E-02	1.437	[Mthfd1]
GO:0019346	transsulfuration	1.00	3.658E-02	1.437	[Mthfd1]

Comparison of quadriceps whole-muscle O-GlcNAcome with whole mouse heart O-GlcNAcome

Recently, whole mouse hearts were studied for O-GlcNAc using a set of “PTMScan” antibodies to directly enrich glycoproteins (Fig. 4a) (Burt et al. 2021). Narayanan et al identify over 1,000 hits, with 34% of them being previously annotated O-GlcNAc proteins and 77% of them being nuclear, cytosolic, and mitochondrial proteins, the major sites of O-GlcNAc proteins (Fig. 4a) (Narayanan et al. 2023). Heart muscle and quadriceps are likely to share conserved hits, but are also expected to contain differences between these distinct organs. During our comparison, we observed 57% overlap between the glycoproteins we enriched from mouse quadricep skeletal muscle proteins and the mouse heart proteins in the Narayanan et al study (Fig. 4b). The full list is show in Supplementary Table 2.

Analysis of the GO terms showed key similarities including canonical glycolysis, long-chain fatty acid import, and aerobic electron transport gene. However, there were many differences in the heart dataset that were not observed in skeletal muscle, such as gamma-aminobutyric acid (GABA) metabolism, vasoconstriction of arteries, and 2-oxoglutarate metabolism. These distinct processes may arise from neurons and adipocytes that were isolated in the hearts alongside cardiac muscle. Other GO terms observed in our skeletal muscle analysis, but not the heart study, were glutamine catabolism, skeletal muscle contraction (Fig. 4c). The full list of heart GO processes is shown in Supplementary Table 3 and visualized as network diagrams, pie chart, and number of genes in Supplementary Figs. 6–8.

In this comparison, we found that many proteins were conserved, with approximately 57% overlap between our quadriceps glycoproteins and proteins in the whole heart dataset (Narayanan et al. 2023). The top gene ontology terms primarily represented functions and processes related to metabolic regulation, a known function of O-GlcNAc effects in cells. In both tissues, Canonical Glycolysis was the top hit, with key glycolytic proteins ALDOA, ENO3, 6-phopshpructokinase, muscle type (PFKM), phosphoglycerate mutase 2 (PGAM2), phosphoglycerate kinase 1 (PGK1), and TPI1 conserved in both tissue types. Electron transport chain processes ranked highly in both datasets, with the coenzyme Q9 homolog, mitochondrial (COQ9), dihydrolipoamide dehydrogenase (DLD), NADH:ubiquinone oxidoreductase subunit A10 (NDUFA10), and protein deglycase (DJ-1) proteins found in both tissues. Besides sugar metabolism, Long-Chain Fatty Acid Import was also conserved, with glutamyl-prolyl-tRNA synthetase 1 (EPRS) and FABP3 glycoproteins being shared in both tissues. However, this fatty acid metabolism GO term ranked highly in quadriceps and lower in heart, potentially reflecting some differences in tissue-specific O-GlcNAc activity on these proteins.

Other tissue-specific differences were found, including skeletal muscle contraction (GO term #9 in quadriceps) vs. vasoconstriction of arteries (GO term #7 in hearts). Organic acid metabolism was also slightly different, with glutamine catabolism processes ranking stronger in quadriceps and GABA and 2-oxoglutarate metabolism being top GO terms in the heart. Finally, we saw specific tissue protein isoform differences, such as skeletal muscle alpha-actin (ACTA1) in quadriceps vs. the cardiac alpha-actin (ACTC1) isoform in hearts.

Besides tissue isoform differences, other sources of variable proteins could arise from non-muscle proteins that would have been isolated alongside the quadriceps or heart, such as fatty tissue, neurons, or vasculature. The methods used between each study were also different, which may lead to nonspecific enrichments tied to the antibody strategy used for the heart proteins vs. the streptavidin strategy used for the quadriceps proteins. Finally, we note that in the heart study a relatively lower set of proteins is known to be O-GlcNAcylated (34%), but the authors estimated that 77% of them were consistent with localization of O-GlcNAc proteins (Narayanan et al. 2023). Overall, we feel that there is strong overlap between quadriceps proteins enriched in this study and the heart proteins that reflect similarities in muscle tissue. We also noted differences between the two tissue types that reflect different proteins expressed in skeletal vs. cardiac muscle, so these types of reference datasets can be useful for predicting glycobiology effects in different organs.

Discussion

Protein O-GlcNAcylation is a key intracellular nutrient sensing process that regulates metabolism and signaling (Wells et al. 2003; Hart 2019). Collectively, O-GlcNAcylation is annotated on nearly 8,000 human proteins, putting these intracellular glycosylation events on par with phosphorylation, ca. 13,000 human proteins, (Vlastaridis et al. 2017) as one of the most ubiquitous protein modifications in cells. Typical O-GlcNAc labeling experiments in cells reveal hundreds of glycosylated proteins, (Li et al. 2019) with just over 2,000 proteins being the highest number of labeled proteins achieved in a single experiment (Woo et al. 2018). Unlike cellular studies, high coverage O-GlcNAc profiles have only been collected from a select few tissue types, specifically just the brain, (Trinidad et al. 2012; Wang et al. 2017; Burt et al. 2021) heart, (Narayanan et al. 2023) and placenta (Luna et al. 2024). For these three tissues, coverage was achieved via O-GlcNAc protein pre-enrichment, revealing up to 1750, 1194, and 452 O-GlcNAc glycoproteins from brain, heart, and placenta, respectively. On the other hand, skeletal muscle datasets from the soleus, cardiac, and gastrocnemius muscles contain just 25 total proteins, (Cieniewski-Bernard et al. 2004; Hédou et al. 2009; Ma et al. 2016) fewer than the hundreds of proteins expected for a metabolically-active tissue (Lambert et al. 2018). These prior skeletal muscle studies used weak lectin affinity chromatography or β-elimination and Michael addition with dithiothreitol (BEMAD) followed by thiosepharose enrichment, which are established but older O-GlcNAc enrichment strategies (Ma and Hart 2017). Therefore, a reference map for O-GlcNAcylated proteins in skeletal muscle was lacking at the onset of the present study.

Skeletal muscle comprises up to 40% of an individual’s body weight, and quadriceps are among the largest muscles in adults. In mice, the quadriceps are the largest muscle by body weight and are similar in anatomy to human quadriceps (Baán et al. 2013). In clinical studies, quadriceps biopsies are performed as a minor surgical procedure to assess insulin resistance in diabetes and cardiovascular disease biomarkers (Barzilay et al. 2009; Doehner et al. 2015). When we were deciding one which tissues to sample for O-GlcNAc profiling, we chose the quadriceps based on their, proclivity to insulin resistance and type 2 diabetes, (Björnholm and Zierath 2005; Karlsson and Zierath 2007; Greene et al. 2018) and potential to be for biopsy samples in clinical applications.

In this study, we used chemoenzymatic O-GlcNAc labeling and biotin enrichment to identify 559 O-GlcNAcylated proteins in the quadriceps skeletal muscle of C57BL/6 mice. Proteins were primarily cytosolic, nuclear, and mitochondrial, matching the expected distribution of O-GlcNAcylated proteins (Wulff-Fuentes et al. 2021). Key GO functional terms that were enriched in the dataset included glucose, fatty acid, amino acid metabolism, as well as regulatory proteins for skeletal muscle contraction, ATP/ADP transport, and cholinergic signaling. When compared to soluble proteins isolated from mouse heart by Narayanan et al, (Narayanan et al. 2023) the majority of the quadriceps O-GlcNAcome overlapped with the heart O-GlcNAcome, with many exceptions being tissue-specific variants of proteins like ACTA1 (α-actin, skeletal muscle) vs. ACTC1 (α-actin, cardiac muscle). Key heart GO functions included GABA metabolism, 2-oxoglutarate metabolism, and arterial vasoconstriction (Narayanan et al. 2023) that were not found in the quadriceps O-GlcNAc functions. Therefore, comparing distinct tissue-specific O-GlcNAc profiles has the potential to identify key tissue-specific physiological and pathophysiological targets for O-GlcNAc-based mechanisms or therapeutics.

Despite recent reviews suggesting O-GlcNAc modification dynamics as a key link between obesity and diabetes, (Ma and Hart 2013; Bolanle and Palmer 2022) very little is known about discrete O-GlcNAc target proteins that are potentially altered in specific tissue types. (Gonzalez-Rellan et al. 2022) This study was performed only in mid-life lean, wild-type C57BL/6 mice under routine handling conditions, representing a baseline condition. This work sets the stage for future studies comparing effects of muscle physiopathology (Lambert et al. 2018) metabolic disease, (Gonzalez-Rellan et al. 2022) exercise, and nutrition (Myslicki et al. 2014). This reference dataset of glycoproteins enriched from quadriceps has the potential to fill a key knowledge gap of O-GlcNAc roles in skeletal muscle tissue.

Materials and methods

Animals

Male C57BL/6J mice were obtained from Jackson Laboratories (strain 000664). Animals were housed on a 12 h/12 h light/dark cycle with ad libitum water and fed low-fat rodent diet for 6 months. At the end of the study male mice were euthanized via carbon dioxide and their tissues were harvested for glycobiology analyses. All animal experiments were performed in accordance with the protocols approved by the Institutional Animal Care and Use Committee (IACUC) for Wayne State University.

Muscle protein O-GlcNAc chemoenzymatic labeling

Mouse muscle tissues were homogenized in RIPA buffer using a Precellys 24 bead homogenizer (Bertin Technologies) for 3 rounds of 5000 rpm, 5 second bursts. Soluble proteins were isolated by centrifuging the samples for 7,000 × g for 10 min at 4 °C, collecting the supernatant, and a second spin at 10,000 × g for 15 min at 4 °C. Proteins were extracted in RIPA buffer (10 mM Tris pH 7.2, 150 mM NaCl, 0.4% sodium dodecyl sulfate, 2% triton X-100, and 1% deoxycholine) with protease inhibitor cocktail (Thermo Fisher, cat. Number J61473-XF). Muscle protein extracts were subjected to chemoenzymatic O-GlcNAc glycoprotein labeling following the methods describe by the Hsieh-Wilson group. (Thompson et al. 2018) Briefly, the engineered Y289L mutant bovine galactosyltransferase (GalT) enzyme was produced in Escherichia coli strain BL21(DE3) following the reported method (Thompson et al. 2018) with just one modification, where bacteria were lysed using a French press device rather than by sonication. Our purified GalT matched the expected size by SDS-PAGE analysis and was sufficiently pure (Supplementary Fig. 1). Positive and negative controls using the known O-GlcNAcylated protein α-crystallin confirmed activity of the construct. Muscle protein extracts were incubated with GalT and the azide-labeled uridine diphosphate N-acetylgalactosamine donor sugar (UDP-GalNAz, Thermo Fisher, cat. Number NC2034688) to selectively install an azide tag on O-GlcNAc sites. The reported methods were followed exactly, substituting our muscle protein extracts instead of protein extracts from cell culture as described in the former report. (Thompson et al. 2018) Following O-GlcNAc chemoenzymatic azide-tagging, the tagged proteins were labeled with biotin-alkyne (Click Chemistry Tools, cat. Number CCT-1266) following the reported method with no changes. (Thompson et al. 2018) Labeled proteins were precipitated from methanol/chloroform/H2O to remove unreacted alkyne reagents. Protein pellets were stored at −80 °C until proteomic preparation.

O-GlcNAc protein enrichment and digestion

Following biotin labeling, proteins were resuspended in 500 μL RIPA buffer (150 mM NaCl, 10 mM tris, pH 7.2, 0.4% SDS, 2% Triton X-100, 1% deoxycholate (DOC)). To enrich biotinylated proteins, samples containing 400 μg total protein were incubated with streptavidin-coated magnetic beads (New England Biolabs, cat. Number S1420S) overnight at 4 °C. Beads were washed with RIPA buffer, wash buffer (50 mM Tris, pH 7.2, 2% SDS), and twice more with RIPA buffer. The beads were then resuspended in DTT in PBS and were then treated with iodoacetamide. The beads were then washed with mass spectrometry (MS) grade water and 50% acetonitrile/50% water, then were resuspended in MS grade 50% acetonitrile/50% water. The samples were digested with Lys-C protease (New England Biolabs cat. Number P8109S) for 16 h at 37 °C and with Solu-Trypsin (Millipore Sigma cat. Number EMS0004) for 1 h (47 °C) then for 4 h (37 °C). The supernatants were quenched with formic acid and the magnetic beads were removed. Samples were dried via vacuum centrifugation and stored at −80 °C.

Proteomics liquid chromatography–tandem mass spectrometry (LC–MS/MS) analysis

Each peptide sample was solubilized in 0.1% trifluoroacetic acid, purified by solid-phase extraction (C18 ZipTip, Millipore, Billerica, MA) and subjected to high-performance liquid chromatography-electrospray ionization-tandem mass spectrometry (HPLC-ESI-MS/MS) analysis. HPLC-ESI-MS/MS was performed on a Thermo Scientific™ Orbitrap Fusion™ Lumos™ Tribrid™ mass spectrometer, fitted with a nanospray flex Ion source (Thermo Fisher, San Jose, CA). Online HPLC was performed using an Thermo Scientific™ Dionex™ UltiMate 3000 RSLC nano-System: NCS-3500RS, with a C18-reversed phase column (75 μm ID, 45 cm length) packed in-house with ReproSil-Pur C18-3 μm resin (Dr. Maisch GmbH, Germany), maintained at 60 °C, connected to non-coated emitter. The peptides were separated with a binary buffer system of 0.1% FA in water (buffer A) and 0.1% FA in acetonitrile (buffer B), at a constant flow rate of 300 nL/min. The peptides were separated with a linear gradient chromatography of 2%–35% buffer B (100% ACN and 0.1% FA in H2O) over a 230 min in-house method; starting with 2%B and increase to 5% at 8 min, to 10% B in 50 min, 25% B in 100 min, to 35% B in 80 min, to 90% B in 1 min, hold at 90% B for 1 min, and then 2% buffer B at a flow rate of 300 nL/min.

A “top 3 second” data-dependent tandem mass spectrometry approach was utilized to identify peptides in the samples. In a top 3 s scan protocol, a full MS1 scan spectrum (survey scan, 400–1600 Th) is acquired using orbitrap mass analyzer followed by collision-induced dissociation (CID) mass spectra MS2, obtained using linear ion trap. The survey scan was acquired using the Orbitrap mass analyzer to obtain high mass accuracy and high mass resolution data (120,000 resolution and mass tolerance of 10 ppm), and top intensity ions were selected for fragmentation during the 3 s time window, using a 1.3 m/z isolation window and a CID collision energy of 35%. Dynamic exclusion was set at 30 seconds. The charge state rejection function was enabled with “unassigned” and “single” charge states rejected. By knowing the accurate mass and fragmentation pattern of the peptide, the peptide’s amino acid sequence can be reliably inferred.

Proteomics data analysis

All mass spectra were analyzed with MaxQuant (Cox and Mann 2008) software version 2.0.03. (Tyanova et al. 2016) The RAW mass spectrometry files were searched against a database with forward and reversed Mus musculus Uniprot protein sequences downloaded from www.uniprot.org. Standard settings in the MaxQuant were applied. Parent mass tolerance was 5 parts per million (p.p.m.), and fragment mass tolerance was 0.5 Da. Two missing trypsin cleavage sites were allowed. Methionine oxidation (M), phosphorylation (STY), and acetyl (protein N-term) were allowed as variable modifications. Carbamidomethyl (C) as fixed modification. The false discovery rate (FDR) for proteins and peptides (with minimum seven amino acids) was set to 0.01. The full, processed data are presented in Appendix Table 1.

Gene ontology (GO), interactome, localization, and O-GlcNAcome overlap analyses

Gene ontology (GO) was conducted with CLUE-GO (Bindea et al. 2009) in Cytoscape. (Shannon et al. 2003) The functional analysis mode was used with “GO: biological processes” with “Evidence: All_Experimental” as ontology terms, with Network Specificity set to “Detailed.” Interactome and protein localization analyses were performed with the StringApp (Doncheva et al. 2019) using the STRING-db (Szklarczyk et al. 2019) dataset in Cytoscape. (Shannon et al. 2003). Enrichment and over-representation test were applied to annotate the functional differences between labeled proteomics data in PANTHER. (Mi et al. 2019) Plots were generated in Perseus and PRISM. Databases used in this study include UniprotKB (https://www.uniprot.org), the O-GlcNAc Database (https://www.oglcnac.mcw.edu) and the O-GlcNAcATLAS (https://www.oglcnac.org/atlas). O-GlcNAc overlap with performed by comparing the UniprotIDs in the O-GlcNAc Database and the O-GlcNAcATLAS with the majority UniprotID’s for MS hits.

Western blotting

To analyze muscle protein extracts via immunoblot, extracts in RIPA buffer were centrifuged (12,000 × g for 10 min at 4 °C) to collect the soluble protein fraction. Protein concentrations were determined via Pierce Rapid Gold BCA Protein Assay Kit (Thermo Fisher, cat. Number A53225). Samples were boiled in SDS gel-loading buffer for 5 min. Proteins were separated on a 4%–12% gradient gel (NuPAGE™ 4% to 12%, Bis-Tris, 1.0–1.5 mm, Mini Protein Gels; Thermo Fisher cat. Number NP0321BOX) and transferred to a nitrocellulose membrane (iBlot™ 2 Transfer Stacks, nitrocellulose; Thermo Fisher cat. Number IB23002) using an iBlot 2 dry blotting system (Thermo Fisher). After blocking with 5% (w/v) milk powder or 5% bovine serum albumin (Research Products International, catalog number A30075) in PBST buffer (phosphate bufgered saline (PBS), 0.05% Tween 20) for 1 h, the membrane was incubated with the appropriate antibody following the manufacture’s protocol. The signals from the antibodies were detected via iBright™ FL1500 instrument (Thermo Fisher).

To detect biotinylated proteins, the membranes were incubated with horseradish peroxidase (HRP)-conjugated streptavidin and visualized with SuperSignal West Pico chemiluminescent substrate (Thermo Fisher, cat. Number 34580).

To confirm hits in the O-GlcNAcylated pulldown samples, the chemoenzymatic O-GlcNAc labeling was repeated on a new set of 500 ug of tissue extracts for each mouse quadricep (n = 3). Following biotin labeling, 200 ug of labeled samples were enriched on 50 uL of streptavidin beads at 4 °C overnight. The next day, samples were washed with RIPA buffer and directly eluted into 40 uL of 1x SDS loading buffer. Western blots for the target proteins fatty acid binding protein 3 (FABP3, Cell Signaling cat. Number 14,780), oxoglutarate dehydrogenase (OGDH, Cell Signaling cat. Number 26,865), glycogen synthase 1 (GYS1, Cell Signaling cat. Number 3,893), Rad23 homolog B (RAD23B, Cell Signaling cat. Number 24,555), and calmodulin 3 (CALM3, Cell Signaling 35,944). Targets were probed with horseradish peroxidase secondary enzymes and detected with SuperSignal West Atto Ultimate Sensitivity chemiluminescent substrate (ThermoFisher cat. Number A38,554). Full western blot images are shown in Supplementary Fig. 9.

Software, databases, and data repository

Statistics and graphics were performed in GraphPad Prism. Proteomic data are available on the public repository under the reference PDX060554. Figures were created with Biorender and ChemDraw.

Author contributions

Ruchi Jaiswal (Data curation [supporting], Investigation [lead], Methodology [lead], Writing—review & editing [supporting]), Yimin Liu (Data curation [lead], Investigation [lead], Methodology [lead], Writing—review & editing [supporting]), Xiangmin Zhang (Data curation [supporting], Investigation [supporting], Writing—review & editing [supporting]), Michael Petriello (Funding acquisition [supporting], Investigation [supporting], Project administration [supporting], Writing—review & editing [supporting]), Zhengping Yi (Conceptualization [equal], Data curation [equal], Funding acquisition [equal], Project administration [equal], Writing—original draft [equal]), Charlie Fehl (Conceptualization [lead], Funding acquisition [lead], Project administration [lead], Visualization [lead], Writing—original draft [lead]).

Funding

This work was supported by the following funding sources: NIH/NIGMS R35GM142637 to C.F. (reagents), Mizutani Foundation for Glycosciences Research Grant 230026 (PI: Fehl), NSF CAREER award CHE-2235508 to C.F. (salary support for Y.L.); NIH/NIEHS R01ES035692-01A1 to M.P. (animal studies), NIH/NIDDK R01 DK128937 to Z.Y. (proteomics data acquisition), and P30ES036084 (support of M.P.).

Conflict of interest: The authors have no conflicts of interest to declare.

Data availability

The data underlying this article are available in the article and in its online supplementary material. The proteomics.RAW files are available on the PRIDE repository with accession numbers PDX060554.