Anomeric fatty-acid functionalization prevents non-enzymatic S-glycosylation by monosaccharide metabolic chemical reporters

Abstract

Metabolic chemical reports have fundamentally changed the way researchers study glycosylation. However, when administered as per-O-acetylated sugars, reporter molecules can participate in non-specific chemical labeling of cysteine residues termed S-glycosylation. Without detailed proteomic analyses these labeling events can be indistinguishable from bona fide enzymatic labeling convoluting experimental results. Here, we report a solution in the synthesis and characterization of two reporter molecules functionalized at the anomeric position with hexanoic acid: 1-Hex-GlcNAlk and 1-Hex-6AzGlcNAc. Both reporters exhibit robust labeling over background with negligible amounts of non-specific chemical labeling in cell lysates. This strategy serves as a template for the design of future reporter molecules allowing for more reliable interpretation of results.

INTRODUCTION

Glycosylation is a family of co- and post-translational modifications (PTMs) that describe the addition of one or more monosaccharides onto specific protein residues. Most protein glycosylation events fall into one of three subgroups: N-linked glycosylation, mucin O-linked glycosylation, and O-GlcNAc modifications.¹ Glycoproteins are present in all cells, with functions including immune recognition,² protein stability,³ cell signaling,^4,5 and cellular trafficking.⁶

Despite a clear link between proper glycoprotein dynamics and cell development, glycosylation is challenging to study due to the intrinsic heterogeneity associated with oligosaccharide motifs.^1,7 Fortunately, metabolic chemical reporters (MCRs) have emerged as powerful tools for studying glycosylation and glycoproteins.⁸ MCRs work by exploiting endogenous monosaccharide salvage pathways present in all eukaryotic organisms.⁹

With this strategy, researchers can introduce small abiotic functional groups such as azides and alkynes that can participate in bioorthogonal reactions to introduce tags that allow for the identification and visualization of glycoproteins in biologically relevant contexts (Fig. S1).^10,11

Carbohydrate MCRs are introduced to cells as per-O-acetylated monosaccharides allowing for their passive diffusion across the cell membrane (Fig. 1a).^12,13 The O-acetyl groups can then be removed by endogenous esterase enzymes to release a free sugar ready to enter its associated salvage pathway. From here, MCRs are metabolized into high-energy nucleotide diphosphate (NDP) sugar donors that serve as substrates for various glycosyltransferases generating labeled glycoproteins.¹⁴ Acetyl groups mask polar hydroxyls which greatly enhances cellular uptake allowing for robust labeling with relatively low treatment concentrations. This practice remained unquestioned until a recent study by the Chen lab reported that treatment with per-O-acetylated MCRs resulted in a significant number of chemically labeled cysteine residues in a process termed S-glycosylation (Fig. 1a).¹⁵ A follow-up mechanistic investigation showed that non-specific chemical labeling follows an elimination-addition reaction. Once the anomeric acetyl is removed, monosaccharides exist in an equilibrium between their linear and ring conformations (Fig. 1a and Fig. S3). When linear, the presence of acetyl groups at the C3 and C4 positions facilitate a β-elimination reaction susceptible to addition from endogenous thiols (Fig. S3).¹⁶ The resulting covalent adducts are indistinguishable from enzymatic labeling events without advanced mass-spectrometry (MS) analysis, convoluting results from routine labeling experiments. This background labeling seems to be context dependent, as other labs have found that MCR labeling can be efficiently blocked by glycosyltransferase inhibitors¹⁷ or removed with glycosidases.¹⁸ Furthermore the Chen lab themselves showed that RNAi knockdown of OGT essentially eliminated 6AzGlcNAc labeling, indicating that the vast majority of the modification is due to enzyme activity.¹⁹

Figure 1.

Per-O-acetylated sugars can chemically modify cysteine residues. (a) Per-O-acetylated sugars passively diffuse across the cell membrane where they are deacetylated by esterase enzymes. (Top) Endogenous salvage pathway enzymes metabolize sugar molecules to corresponding UDP sugar donors that enzymatically label cells. A detailed scheme of GlcNAc salvage pathway metabolism is shown in Fig. S2. (Bottom) Once the anomeric acetate is removed, sugars exist in an equilibrium between linear and ring conformations. In its linear confirmation, partially acetylated molecules can be covalently added to native cysteine residues resulting in non-specific chemical labeling. A detailed scheme of this mechanism can be found in Fig. S3. (b) General synthetic approach to synthesize fatty acid functionalized MCRs. (1) Per-O-acetylated sugars form a stable oxazoline intermediate. (2) C3, C4, and C6 acetates are deprotected. (3) Carboxylic acids are coupled, opening the oxazoline and forming fatty acid functionalized sugar molecules.

The specific conditions and cell or tissue types that result in significant background labeling need to be clarified, but at the same the Chen lab has made progress towards the preparation of MCRs that display less cysteine modification. Specifically, in the case of N-azidoacetyl-galactosamine (GalNAz) they selectively installed propylesters at the 1- and 6-hydroxyl groups.^16,20 However, the selective installation of esters can require multiple steps that may not be universal for different monosaccharide structures. Other past work has demonstrated that synthesizing MCRs with a single large hydrophobic functional group maintains the lipophilicity required for cell permeability.²¹ We rationalized that this approach could be used to circumvent non-specific chemical background labeling. Specifically, we reasoned that designing MCRs with fatty acids on the anomeric position (Fig. 1b) would still be removable by esterases with the resulting product lacking the C3 and C4 acetyl groups necessary to facilitate S-glycosylation. This approach also maintains the freedom to functionalize the other positions with abiotic groups allowing for a continued development of diverse MCRs.

Here, we describe a proof-of-concept study using two well characterized O-GlcNAc MCRs: N-pentynyl-glucosamine (GlcNAlk) and 6-azido-6-deoxy-N-acetyl-glucosamine (6AzGlcNAc).^22,23 We demonstrate that functionalizing both GlcNAlk and 6AzGlcNAc with 6-carbon fatty acids on the anomeric position maintains their ability to label mammalian cells. Importantly, we show that these derivatives have almost undetectable levels of background labeling in cell lysates compared to both their per-O-acetylated and free-OH counterparts. Finally, we confirm the largely enzymatic nature of GlcNAlk labeling by showing that expression of a mutant biosynthetic enzyme that can better accommodate the large N-acetyl-group, AGX1(F383G), dramatically improves the labeling efficiency. Taken together, we believe that this approach can be widely adopted to circumvent chemical labeling in experiments using MCRs.

RESULTS AND DISCUSSION

Fatty acid functionalized metabolic chemical reporters label proteins in mammalian cells.

To test the labeling efficiency of fatty acid functionalized MCRs, we synthesized a series of GlcNAlk derivatives with acyl chains increasing in length from four to six carbons (4–6, Fig. 2a). HeLa cells were metabolically labeled with 200 μM of each reporter for 72 h. Cells were then harvested, washed, and lysed. Soluble protein fractions were reacted with an azide functionalized TAMRA dye (Az-TAMRA) using copper-catalyzed azide-alkyne cycloaddition (CuAAC). In-gel fluorescence scanning showed successful labeling of 4–6 with increasing efficiency corresponding to increasing chain length of the fatty acid substituent (Fig. 2b). 1-Hex-GlcNAlk (6) labels cells at the highest level. This is consistent with previous reports²¹ which demonstrate increasing the lipophilic character results in more efficient diffusion across the cell membrane. Importantly, all three qualitatively label the same subset of proteins consistent with each being converted into the same metabolic substrate; UDP-GlcNAlk. To test the concentration dependence of 6, HeLa cells were treated with various concentrations to 200 μM for 72 h. Cells were lysed and subjected to CuAAC with Az-TAMRA (Fig. 2c). In-gel fluorescence scanning revealed that labeling with 6 starts with treatment concentrations as low as 25 μM and increases up to 200 μM. Next, we examined the kinetics of protein labeling. HeLa cells were treated with 6 (200 μM) for different lengths of time followed by CuAAC and analysis by in-gel fluorescence (Fig. 2d). Labeling was detectable at 12 h with marginal increases up to 72 h.

Figure 2.

Characterization of GlcNAlk derivatives. (a) Structures of GlcNAlk derivatives (b) Labeling efficiency for GlcNAlk derivatives increases with increasing fatty acid chain length; HeLa cells were treated with 200 μM of indicated MCR for 72 h (c) 1-Hex-GlcNAlk labeling is concentration dependent. HeLa cells were treated with 1-Hex-GlcNAlk at concentrations from 0 – 200 μM for 72 h (d) 1-Hex-GlcNAlk labeling is time dependent. HeLa cells were treated with 1-Hex-GlcNAlk (200 μM) for 0 – 72 h.

After determining that fatty acid functionalization of GlcNAlk results in robust labeling, we wanted to confirm that this approach can be broadly applied with other known MCRs. We synthesized a hexanoic acid analog of the O-GlcNAc selective MCR 6AzGlcNAc termed 1-Hex-6AzGlcNAc (11). 1-Hex-6AzGlcNAc was characterized in HeLa cells. First, HeLa cells were treated with 11 for different lengths of time followed by lysis and CuAAC with an alkyne functionalized TAMRA tag (Alk-TAMRA) for analysis using in-gel fluorescence scanning (Fig. S4a). The results of this experiment demonstrate that labeling was detectable at 12 h and peaked between 24 h and 48 h. In the case of 1-Hex-6AzGlcNAc, we observed some cycling of the labeling signal at days 2 and 3 of treatment. We hypothesize that this may result from the removal of 6AzGlcNAc from the protein substrate, its relatively slow metabolism back to UDP-6AzGlcNAc and re-addition to proteins by OGT. We have observed this subtle cycling in the past with GalNAz during a chain experiment.²⁴ We then performed a concentration course by treating HeLa cells with various concentrations of 11 for 24 h (Fig. S4b). Analysis by in-gel fluorescence show labeling efficiency is dose-dependent, a characteristic consistent with other MCRs. Importantly, we did not observe any notable toxicity from any of the treatment conditions.

To explore the effect of different cellular metabolisms on labeling efficiency, we treated a panel of cell lines (Fig S5a and b). In addition to HeLa cells, NIH3T3 cells and CHO cells were treated with 200 μM of 6 or 11 for 72 and 24 h respectively. Cells were all lysed and subjected to CuAAC with either Az-TAMRA or Alk-TAMRA and analyzed via in-gel fluorescence scanning. There was robust labeling over background for both reporters in most cell lines, the one exception being fairly low labeling over background by 1-Hex-6AzGlcNAc in NIH3T3 cells. A qualitative evaluation of the banding patterns showed a diverse pattern and intensity of modified proteins for both reporters consistent with different expression profiles in each cell line. Notably, labeling efficiency in NIH3T3 cells was the lowest for both reporters whereas 6 labels HeLa cells the highest over background and 11 labels CHO cells the highest over background.

Finally, we tested the metabolic labeling efficiency of 6 and 11 compared to fully unprotected GlcNAlk and 6AzGlcNAc respectively (Fig. S6a and b). HeLa cells were treated with 200 μM of the indicated reporter followed by lysis and CuAAC. Labeled proteins were visualized with in-gel fluorescence scanning. Both 6 and 11 label more efficiently than their corresponding free sugar. However, qualitatively they labeled the same pattern of proteins. These results demonstrate the importance maintaining a ratio of hydrophilicity and lipophilicity to passively diffuse across the cell membrane and efficiently label cells but also serve as validation that the MCR-labeled proteins are enzymatic substrates as the free sugars do not participate in non-enzymatic background labeling.¹⁵

Replacement of per-O-acetyl protecting groups with an 1-O-hexanoic ester prevents non-enzymatic S-glycosylation.

To access the potential for 1-Hex-GlcNAlk (6) and 1-Hex-6AzGlcNAc (11) to participate in non-specific chemical modification of cysteine residues, we incubated these compounds at concentrations of 0.2 mM or 2 mM in HeLa cell lysates under the conditions reported to result in S-glycosylation.¹⁵ For comparison, HeLa cell lysates were also incubated with free sugar (GlcNAlk and 6AzGlcNAc) and per-O-acetylated sugar (Ac₄GlcNAlk and Ac₃6AzGlcNAc) (Fig. 3a–d). All conditions were subjected to CuAAC with either Az-TAMRA in lysates treated with alkyne-bearing probes or Alk-TAMRA in lysates treated with azide-bearing probes. HeLa cells incubated with per-O-acetylated reporters resulted in robust labeling over background, consistent with the results of previous studies investigating S-glycosylation.¹⁵ However, HeLa cells incubated with hexanoic functionalized reporters demonstrate negligible lysate labeling supporting a model where substituents on the anomeric position of carbohydrate MCRs do not facilitate the elimination-addition mechanism necessary for S-glycosylation to proceed.

Figure 3.

Neither 1-Hex-GlcNAlk or 1-Hex-6AzGlcNAc chemically modify cell lysates. (a) Structures of GlcNAlk derivatives used for experiments (b) Non-denatured HeLa cell lysate was treated with GlcNAlk, Ac₄GlcNAlk, or 1-Hex-GlcNAlk (c) Structures of 6AzGlcNAc derivatives used for experiments (d) Non-denatured HeLa cell lysate was treated with 6AzGlcNAc, Ac₄6AzGlcNAc, or 1-Hex-6AzGlcNAc.

Notably, lysates incubated with free sugars result in higher background labeling than the corresponding hexanoic derivatives. We attribute this labeling to separate non-enzymatic covalent modification through glycation. Glycation occurs between the aldehyde present in the open sugar conformation and nucleophilic residues such as lysine and has been purposely detected using azide-containing free sugars.²⁵ In cell lysates, free sugars can exist in an equilibrium between linear and ring confirmations and therefore can readily participate in glycation events. Conversely, the hexanoic acid protecting the 1-hydroxyl mitigates any potential background glycation in the 1-Hex-MCR derivatives.

1-Hex-GlcNAlk labeling increases in cells expressing mutant AGX1(F383G).

We next compared Ac₄GlcNAlk to 1-Hex-GlcNAlk and found that the 1-Hex-MCR labeled proteins with significantly less intensity (Fig. S7). We reasoned that this difference in labeling efficiency could be due to less diffusion of the 1-Hex compound into cells and/or that a large amount of the observed Ac₄GlcNAlk labeling results from background S-glycosylation in HeLa cells. Carbohydrate MCRs rely on several enzymatic steps to be transformed into the high-energy sugar donor substrates of glycosyltransferases (Fig. S2). After de-acetylation or de-lipidation, GlcNAlk is metabolized into GlcNAlk-1-phosphate and then converted to UDP-GlcNAlk by the enzyme AGX1. Previous work demonstrates that unnatural sugars with bulky substituents at the N-acetyl position are poor substrates for both isoforms of UDP-GlcNAc pyrophosphorylase (AGX1/2), the enzyme responsible for converting GlcNAc-1-P to UDP-GlcNAc in the GlcNAc salvage pathway (Fig. S2).²⁶ The X-ray crystal structure of human AGX1 verifies that the N-acetyl group of GlcNAc-1-P resides in a compact hydrophobic pocket. Thus, mutation of this pocket to yield AGX1(F383G) results in more efficient metabolism of certain MCRs.²² We therefore decided to use AGX1(F383G) expression to determine whether 1-Hex-GlcNAlk enters cells inefficiently or is simply metabolized slowly to UDP-GlcNAk. If the bottleneck is largely metabolic, we reasoned that AGX1(F383G) would result in improved synthesis of UDP-GlcNAlk and higher protein labeling by glycosyltransferases. Accordingly, HeLa and NIH3T3 cells expressing AGX1(F383G) were treated with 200 μM of 6 for 72 h (Fig. 4a and b). Cells were then lysed and reacted with Az-TAMRA using CuAAC. Analysis with in-gel fluorescence scanning shows a large increase in protein labeling compared to their wild-type counterpart strongly suggesting that our 1-Hex-MCRs efficiently diffuse into cells and largely result in enzyme-dependent modification of proteins. Furthermore, it suggests that a notable fraction of Ac4GlcNAlk signal results from S-glycosylation in HeLa cells, as result that is consistent with our prior observation showing poor GlcNAlk metabolism is these cells.²⁷

Figure 4.

1-Hex-GlcNAlk labeling increases in cells expressing mutant AGX1. NIH3T3 (a) and HeLa (b) cells expressing AGX1(F383G) show higher labeling efficiency when treated with 1-Hex-GlcNAlk than those expressing endogenous wild-type AGX1.

CONCLUSION

The discovery that per-O-acetylated MCRs have the potential to result in chemical labeling in a process termed S-glycosylation calls in to question the validity of conclusions drawn from the use of these tools. Work toward understanding the mechanism indicates that acetate groups on the C3 and C4 position facilitates the elimination-addition reaction thought to drive the covalent bond formation between endogenous cysteine residues and unnatural sugar molecules resulting in adducts indistinguishable enzymatic labeling events in many routine experiments.

Here, we propose a solution in the design and synthesis of MCRs functionalized with hexanoic acid at the anomeric position. We reasoned that a single long acyl chain would maintain the lipophilicity necessary for passive diffusion across the cell membrane while remaining a substrate for esterase enzymes to cleave and release an active molecule. To test this, we synthesized two MCRs: 1-Hex-GlcNAlk and 1-Hex-6AzGlcNAc. When characterized in HeLa cells, both exhibited robust labeling over background in a concentration and time dependent manner (Fig. 2 and Fig. S4). Importantly, cell lysates incubated with these reporters show negligible non-specific chemical labeling validating our hypothesis (Fig. 3). Finally, we found that AGX1(F383G) expression dramatically improves the modification of proteins upon 1-Hex-GlcNAlk treatment (Fig 4). Because wild-type AGX1 is a known bottleneck for the transformation of N-acetyl-modified MCRs to their corresponding UDP-sugar donors, this result indicates that our labeling is overwhelmingly due to enzymatic protein modification. The results of this study nicely complement the work of the Chen lab and others and offer an alternative approach to the design and synthesis of future carbohydrate MCRs that avoid confounding background chemical-modification of cysteine residues.

DATA AVAILABILITY

All data are contained in this article.