Monitoring Protein O-GlcNAc Status via Metabolic Labeling and Copper-free Click Chemistry

Abstract

O-GlcNAc modification found on the serine and threonine residues of intracellular proteins is an inducible post-translational modification that regulates numerous biological processes. In combination with other cell biological and biochemical approaches, a robust and streamlined strategy for detecting the number and stoichiometry of O-GlcNAc modification can provide valuable insights for decoding the functions of O-GlcNAc at the molecular level. Herein, we report an optimized workflow for evaluating the O-GlcNAc status of proteins using a combination of metabolic labeling and click chemistry based mass tagging. This method is strategically complementary to the chemoenzymatic-based mass-tagging method.

Introduction

Intracellular O-linked β-N-acetylglucosamine (O-GlcNAc) is a dynamic post-translational modification that displays extensive crosstalk with serine and threonine phosphorylation (1). The reversible nature of O-GlcNAcylation is achieved by a pair of O-GlcNAc cycling enzymes, OGT and OGA, for the addition and removal of the carbohydrate moiety, respectively. The degree of O-GlcNAc modification varies from site to site and protein to protein, frequently in response to intracellular and extracellular stimuli. An elegant mass-tagging strategy to assess O-GlcNAc stoichiometry was reported previously by Hsieh-Wilson and coworkers using a combination of chemoenzymatic labeling and aminooxy-mediated click reaction (2,3). In their approach, a keto-functionalized donor substrate (UDP-keto-Gal) is added enzymatically onto O-GlcNAc modified proteins using a genetically engineered β-1,4-galactosyltransferase, Y289LGalT. Subsequently, an aminooxy-group containing PEG derivative, with a mass of either 2 kDa or 5 kDa, is added to the reaction mixture containing keto-decorated O-GlcNAc-modified proteins. The proportion of O-GlcNAc-modified proteins to their unmodified counterparts can be detected by the classical immunoblotting procedure using an antibody against the protein of interest after separating the unmodified and mass-tagged proteins on a one-dimensional SDS-PAGE gel. Importantly, the application of this chemoenzymatic-based mass-tagging approach is restricted to in vitro manipulation and the introduction of the keto group is rather laborious and time-consuming. Since O-GlcNAc research relies heavily on samples obtained from cell culture preparations, we reasoned that a modified mass-tagging workflow with metabolic incorporation of a clickable bioorthogonal group could be an important addition to the O-GlcNAc toolkit.

Toward this end, we utilized a metabolic labeling approach to introduce an azido functionality onto O-GlcNAc modified proteins by feeding HEK293T cells with 25 mM of peracetylated azidogalactosamine (Ac₄GalNAz) for 48 hours. As demonstrated previously by Bertozzi’s group, Ac₄GalNAz can be catabolized via the GalNAc salvage pathway followed by a conversion through GALE epimerase to generate UDP-GlcNAz (4), which can be recognized by the mammalian OGT as a donor substrate (5). Notably, Ac₄GalNAz can also be incorporated into all GlcNAc and GalNAc containing complex O- and N-glycans that are ubiquitously found on the membrane proteins. The presence of these unwanted azido groups in the final samples can be minimized by isolating nuclear or/and cytoplasmic fractions for the click reactions.

For the azido-reactive click functionality, we opted for an azadibenzylcyclooctyne (DBCO, also known as ADIBO) that is designed specifically to react with the azido group via a copper-free strain-promoted azide-alkyne cycloaddition (SPAAC) mechanism (6), since we could easily obtain DBCO-conjugated biotin (DBCO-biotin) or PEG5k (DBCO-PEG5k) from commercial sources. For our initial test, we used 1 mM DBCO-biotin to react with HEK293T nucleocytoplasmic fractions (with or without Ac₄GalNAz labeling) for 20 hours (all reactions were performed at ambient temperature unless stated otherwise), followed by a streptavidin-HRP blotting that recognizes the biotin moiety. Surprisingly, the result revealed comparable levels of biotin incorporation in both samples, suggesting DBCO is highly reactive toward a non-azido functionality that is expressed abundantly in biological systems (Fig. 1a, right). Similar observations have been reported (7,8). For instance, van Geel et al. demonstrated that various cyclooctyne derivatives show strong reactivity toward the sulfyhydryl (or thiol) group on the cysteine side chains (8). In order to prevent the generation of the thiol-DBCO conjugate, we first reduced our samples with 10 mM dithiothreitol (DTT) at 56 °C for 1 hour, immediately followed by an alkylation with 90 mM iodoacetamide (IAA) for 45 minutes in dark. Reduction and alkylation of samples prior to reacting with DBCO-biotin greatly reduces the non-specific reactivity of DBCO in sample without azidosugar labeling (Fig. 1a, left).

Fig. 1.

Optimization of the azido-DBCO click reaction conditions. (a) Alkylation prevents the conjugation between thiol and DBCO functionalities. (b) Azido-DBCO specific reaction reaches completion within one hour. (c) At least 1.6 mM of DBCO reagent per 100 ug of nucleocytoplasmic extract is needed in a typical click reaction.

While alkylation of the cysteine residue improves the selectivity of azido-DBCO reaction, we rationalized that optimizing the duration of the click reaction could further optimize the ratio of the reaction between azido-DBCO functionalities to the formation of thiol-DBCO product. For the time course analysis, DTT- and IAA-treated nucleocytoplasmic samples (with or without Ac₄GalNAz labeling) were incubated in the presence of 1 mM DBCO-biotin for various time periods (ranging from 30 minutes to 16 hours). As shown in Fig. 1b, 30 minute incubation is sufficient for SPAAC as the streptavidin-cross reactivity is detected only in the sample labeled with Ac₄GalNAz. Allowing the reaction to proceed longer than 1 hour leads to the formation of undesired side products in samples devoid of the azidosugar as illustrated by the appearance of streptavidin-cross reactivity (Fig. 1b).

For accurate estimation of O-GlcNAc stoichiometry, it is crucial that excess DBCO is present in the reaction mixture to ensure all the azido functionality in the sample is fully labeled. In order to titrate the amount of DBCO reagent needed for fixed amount of nuclear or cytosolic lysates, we performed a pulse-chase experiment in which samples were first pulsed with various concentration of DBCO-PEG5k and chased by DBCO-biotin. If sufficient DBCO reagent is present in the pulse stage to saturate the azido functionality, no biotin moiety will be added onto the O-GlcNAc modified proteins. Otherwise, the remaining unreacted azido group will be labeled by the DBCO-biotin and can be detected by a streptavidin blot. As a result, a complete labeling of azido functionality at the pulse stage will not result in any streptavidin cross-reactivity. One hundred micrograms of DTT- and IAA- treated nucleocytoplasmic protein lysates were incubated with different concentrations of DBCO-PEG5k for 1 hour (the final concentration of the protein sample in a typical reaction mixture was 1 µg/µl), after which 1 mM of DBCO-biotin was added to the reaction mixture and incubated for another 30 minutes. As illustrated in Fig. 1c, streptavidin-HRP blot shows a robust biotin incorporation in samples treated with less than 1.6 mM of DBCO-PEG5k, indicating the need of 1.6 mM or higher concentration of DBCO to completely react with the available azido groups in 100 µg of nucleocytoplasmic proteins in a final volume of 100 µl.

Collectively, we settled on the following conditions for a typical SPAAC reaction: nuclear/cytoplasmic sample is prepared from cells fed with 25 mM Ac₄GalNAz for 48 hours. After treating the sample with 10 mM DTT and 90 mM IAA, 100 µg of protein is mixed with 2 mM of DBCO-PEG5k in a final protein concentration of 1 µg/µl. The reaction time was for 1 hour.

Having concluded the optimized condition, we next evaluated the stoichiometry of O-GlcNAc on two known O-GlcNAc modified proteins, OGA and Sp1, both of which were also subjected to examination in the initial O-GlcNAc mass-tagging report (3). We detected a small portion of the mono-glycosylated form of OGA and multiple glycosylated forms of Sp1 from the cytoplasmic and nuclear fractions of HEK293T cells (Fig. 2a, densitometry was performed using ImageJ). These results are in good agreement with Rexach et al. (3). Additionally, we also established the presence of monoglycosylated BAT3 (also known as HLA-B-associated transcript 3, Scythe, or BAG-6) using our procedure. BAT3 is a chaperone protein that has been found to be an OGT interacting partner (9). However, whether BAT3 is also an OGT substrate had not been addressed prior to this report. Having shown that BAT3 exists as a mono-glycosylated form, albeit in very low stoichiometry, provides an entry for further investigation of the impact of O-GlcNAc on the regulatory role of BAT3 and exploration of the functional significance of the BAT3-OGT interaction.

Fig. 2.

PEGylation of O-GlcNAc modified proteins. (a) Both OGA and BAT3 are modified by one O-GlcNAc, whereas Sp1 is glycosylated with up to six detectable sites. Denistometry was used to calculate amount of total protein modified. (b) The presence of OGA selective inhibitors (either GNSg or TMG) leads to an increase in the O-GlcNAc modified form of OGA, whereas high glucose treatment does not alter the O-GlcNAc status of OGA. In contrast to OGA, O-GlcNAc status of Sp1 remains similar in all treatments. Densitometry was used to calculate amount of total protein modified.

To demonstrate this workflow can be easily implemented for analyzing multiple conditions at once, we proceeded to compare the O-GlcNAc stoichiometry of OGA and Sp1 in HEK293T cells that were cultured under low or high glucose conditions, as well as in the presence of two different OGA selective inhibitors: GlcNAcstatin-g [GNSg, (10)] and Thiamet-G [TMG, (11)]. As shown in Fig. 2b, we observed that O-GlcNAc stoichiometry of OGA increases in the presence of its own inhibitors (more than 3-fold), yet no significant difference in the O-GlcNAc stoichiometry of Sp1 was detected. The increase in the glycosylated form of OGA upon the inhibition of the hydrolase activity is in agreement with a previous study using PUGNAc, a less selective OGA inhibitor (12). However, the biological significance of this phenomenon remains unclear. The lack of a significant increase in Sp1 glycosylation in a 48 hour labeling experiment suggests that global elevation of O-GlcNAc levels via OGA inhibition is not universal to all modified proteins.

In summary, we described a streamlined and optimized procedure for the measurement of O-GlcNAc stoichiometry using a combination of metabolic labeling and strain-promoted copper-free click chemistry reaction. By introducing the bioorthogonal group in cell culture, our method is strategically complementary to the original approach devised by Hsieh-Wilson and colleagues. Moreover, our workflow omits several onerous hand-on steps in the chemoenzymatic labeling procedure and all needed reagents are commercially available. This procedure is feasible to incorporate into any cell culture based experimental models for O-GlcNAc studies.