Metabolic chemical reporters of glycans exhibit cell-type selective metabolism and glycoprotein labeling

Abstract

Since the pioneering work by Reutter and co-workers that demonstrated structural flexibility in the carbohydrate biosynthesis and glycosylation pathways, many different labs have used unnatural monosaccharide analogs to perform glycan engineering on the surface of living cells. A subset of these unnatural monosaccharides contain bioorthogonal groups that enable the selective installation of visualization or enrichment tags. These metabolic chemical reporters (MCRs) have proven to be powerful for the unbiased identification of glycoproteins; however they do have certain limitations. For example, they are incorporated sub-stoichiometrically into glycans and most MCRs are not selective for one class (e.g., O-GlcNAcylation) of glycoprotein. Here, we explore the relationship between the biosynthesis of MCR donor-sugars in cells and the labeling levels of four different N-acetylglucosamine- and N-acetylgalatosamine-based MCRs. We find that the build-up of the different donor-sugars correlates well with the overall labeling levels but less so with intracellular labeling of proteins by O-GlcNAcylation.

TOC image

Metabolism matters: Metabolic chemical reporters (MCRs) of protein glycosylation are monosaccharide analogs bearing bioorthogonal functional groups that can be used for the installation of tags. Here, we show that cell-lines have different capacities for the metabolism of MCRs and that classes of glycosylation are differentially effected by the build-up of the corresponding donor sugars.

Metabolic chemical reporters (MCRs) of glycans take advantage of the consistently growing number of bioorthogonal reactions for the incorporation of different tags into glycoproteins.^[1–5] Of particular utility to the identification of glycoproteins are MCRs bearing azides or alkynes that can be reacted with enrichment tags ex vivo for subsequent identification using modern proteomics techniques.^[6] Importantly, these MCRs are often easy to prepare chemically and many are commercially available. Additionally, the copper(I)-catalyzed azide-alkyne cycloaddition (CuAAC) employed for tag installation is robust and selective, particularly compared to biological methods like lectin or antibody enrichment. Despite these clear advantages, MCRs do have demonstrated weaknesses compared to methods that detect endogenous carbohydrates. First, they must compete with endogenous monosaccharides and are therefore incorporated into glycans at substoichiometric levels. Second, we and others have demonstrated that many glycoprotein MCRs are not selective for one type of glycosylation but are instead metabolically interconverted and incorporated into multiple classes of glycoproteins.^[7–11] For example, N-azidoacetyl-glucosamine (GlcNAz)^[12] and N-azidoacetyl-galactosamine (GalNAz)^[13] can be metabolized to the corresponding UDP-donors and enzymatically interconverted, resulting in incorporation into intracellular O-GlcNAcylation and cell-surface N-linked and mucin O-linked glycoproteins (Scheme 1).^[8,9,11]

Scheme 1.

Incorporation of metabolic chemical reporters (MCRs) into glycoproteins. Per-O-acetylated analogs of GlcNAc and GalNAc can passively diffuse into cells where they are deacetylated by unknown lipases/esterases. They can then enter the biosynthetic, carbohydrate salvage pathways and be converted to the corresponding uridine diphosphate (UDP) sugar-donors. At this stage, they also have the opportunity to be interconverted by the enzyme UDP-glucose-4-epimerase (GALE), before being incorporated into glycoproteins by a variety of glycosyltransferases.

We have been interested in addressing the selectively of MCRs, with the goal of developing reporters that could be used for the analysis of a single class of glycoproteins. We recently demonstrated that changes to the structure of the MCRs could yield reporters that are highly selective for O-GlcNAcylation, including 6-azido-6-deoxy-N-acetyl-glucosamine (6AzGlcNAc)^[11] and 2-azido-2-deoxy-glucose (2AzGlc).^[14] Here, we begin to explore the relationship between MCR metabolism, leading to MCR-donor-sugar accumulation, and protein incorporation levels. More specifically, we first used synthetic chemistry to prepare the known MCRs GlcNAz and GalNAz and the corresponding alkyne-containing MCRs: GlcNAlk and GalNAlk (Scheme 1), as well as their corresponding UDP-sugar-donors. We then correlated the build-up of the UDP-sugars with the levels of MCR incorporation as determined by fluorescence after CuAAC with an appropriate fluorescence tag. Not surprisingly, we find that the global incorporation of these non-selective MCRs into glycoproteins correlates very well with the levels of the UDP-sugars, both within a cell line and across different cell types. Interestingly, by using trypsin to remove many of the cell-surface proteins prior to cell lysis and CuAAC, we found that intracellular O-GlcNAcylation is much less correlated with UDP-sugar build up. These results indicate that a complicated interplay between metabolism, MCR structure, and enzymatic promiscuity contribute to MCR selectivity and stoichiometry that may change in different cells and tissues.

Figure 1.

Different GlcNAc- and GalNAc-based MCRs label proteins at varying levels in different mammalian cells. The indicated mammalian cells were treated with one of the four MCRs (@ 200 μM concentration) or DMSO vehicle for 16 h. At this time, the cells were collected by physical scraping, lysed, and subjected to CuAAC with either alkynyl- or azido-TAMRA. Labeled proteins were then visualized using in-gel fluorescence scanning. This data is representative of two biological experiments.

To explore the interplay between MCR-structure, metabolism, and labeling efficiency we first synthesized the per-O-acetylated (Ac₄) versions of GlcNAz, GalNAz, GlcNAlk, and GalNAlk. The labeling of cells has been thoroughly characterized for all of these MCRs except for Ac₄GalNAlk. Accordingly, we first treated HeLa cells with increasing amounts of Ac₄GalNAlk to determine the dose-dependence of labeling. After lysis of the cells and CuAAC with azido-TAMRA, the levels of incorporation were visualized by in-gel fluorescence (Supplemental Figure S1A). Consistent with other MCRs, Ac₄GalNAlk showed dose-dependent labeling that continued to increase through our highest concentration of 200 μM. Next, we examined the kinetics of protein labeling by treating HeLa cells with by Ac₄GalNAlk (200 μM) for different lengths of time. Cell lysis, followed by CuAAC and in-gel fluorescence, showed that proteins were labeled by GalNAlk within 2 h and that labeling reached a maximum between 10 and 12 h of treatment (Supplemental Figure S1B). Finally, we investigated the stability of GalNAlk-labeling in a pulse-chase experiment. More specifically, HeLa cells were treated with Ac₄GalNAlk (200 μM) for 16 h, followed by a replacement of the media with fresh media containing Ac₄GalNAc (200 μM). After different lengths of time, the cells were collected and analyzed by CuAAC and in-gel fluorescence (Supplemental Figure S1C). The GalNAlk-dependent signal was lost over the course of 72h, consistent with an MCR that labels a mixture of cell-surface and O-GlcNAcylated proteins.

With GalNAlk-labeling initially characterized, we next individually treated a variety of cell-lines with 200 μM of each of the four different MCRs. The cells were then collected by physical scraping, to preserve as many cell-surface glycoproteins as possible. The corresponding cell lysates were then subjected to CuAAC with an appropriate TAMRA-tag and the global levels of MCR incorporation were visualized by in-gel fluorescence scanning (Figure 1). In general, we observed the strongest labeling by GalNAz, followed by similar levels of incorporation by GlcNAz and GlcNAlk, and finally the lowest levels of labeling by GalNAlk. Next, we chemically synthesized the UDP-sugars corresponding to each of the four MCRs (see Supplemental Information for synthetic details). We first focused on NIH3T3 cells, as they showed relatively high incorporation by all four MCRs. Accordingly, the metabolites from NIH3T3 cells were collected and analyzed the nucleotide sugar content with hydrophilic interaction liquid chromatography (iHILIC) HPLC with UV detection. As can be seen in Figure 2A, no major, endogenous metabolites were detected in untreated NIH3T3 cells between a retention time of 15 and 35 min. However, the synthetic UDP-MCR standards could be readily detected when they were spiked into the metabolite mixture, enabling us to analyze the build-up of these molecules in treated cells. Unfortunately, we were not able to separate the glucose and galactose epimers of the donors and therefore cannot speak to the interconversion of the MCRs. The enzyme that interconverts UDP-GlcNAc and UDP-GalNAc, UDP-glucose 4-epimerase (GALE), is not directionally driven, but instead simple maintains the equilibrium between the two UDP-sugars.^[15] Therefore, the peak is presumably a relatively equal mixture of the UDP-sugars. Notably, we could detect the build-up of UDP-sugars from NIH3T3 cells treated with 200 μM of either Ac₄GlcNAz, Ac₄GalNAz, or Ac₄GlcNAlk; however, no UDP-sugar could be seen from cells treated with Ac₄GalNAlk. We quantitated the levels of UDP-sugar (Figure 3) and found that they correlated quite well with the overall labeling of the different MCRs (Figure 1), suggesting that efficient metabolism is an important factor behind reporter incorporation in this cell line. We next tested two other cell lines, HeLa and HEK293, using the same procedures (Figure 2B&C). Once again, we find that the labeling levels of the different MCRs (Figure 1) correlate well with the amounts of UDP-sugar (Figure 3). Interestingly, however, comparisons across different cell lines do not completely correlate. For example, HEK293 and NIH3T3 cells label a very similar levels when treated with either Ac₄GlcNAz or Ac₄GalNAz despite notably more UDP-sugar build up in NIH3T3 cells. This result can be interpreted in several different ways. For example, it is possible that once a certain threshold of UDP-sugar is present in the cells the amount of MCR incorporation into glycoproteins is “maxed out.” It is also likely, that the levels of endogenous UDP-sugars (e.g., UDP-GlcNAc) are critical, as they will directly compete with MCRs for glycosyltransferases and transporters.

Figure 2.

The UDP-sugars of MCRs build-up to different levels in three cell lines. (A) NIH3T3 cells were treated in triplicate with one of the four MCRs (@ 200 μM concentration) or DMSO vehicle for 6 h before collection of the cellular metabolites by methanol lysis. The corresponding UDP-sugars were then separated on iHILIC-HPLC and detected by UV absorbance (264 nm). DMSO-treated lysates where spiked with synthetic standards of the different UDP-sugars as positive controls. (B & C) HeLa and HEK293 cells were treated and analyzed in the same fashion as NIH3T3 cells.

Figure 3.

Quantitation of UDP-sugar levels after MCR treatment. Peak areas from the HPLC traces in Figure 3 were determined. Error bars represent ±s.e.m. from the mean of biological replicates (n = 3).

In our previous work in this area, we have found somewhat smaller differences in the efficiency of labeling between different MCRs when cells were collected by trypsinolysis, which can remove a large fraction of the cell-surface proteins and thus enrich for labeling of intracellular O-GlcNAcylation. O-Linked-N-acetylglucosamine (O-GlcNAc) transferase (OGT) catalyzes the addition of N-acetylglucosamine to serine or threonine residues of intracellular proteins by using UDP-GlcNAc as a substrate. To determine if labeling of this O-GlcNAcylated fraction also correlates with UDP-sugar levels, the same panel of cells was again individually treated with 200 μM of each of the four MCRs. The cells were then collected by trypsinolysis, lysed, and subjected to CuAAC with the appropriate TAMRA-tag. To confirm that cell-surface glycoproteins were lost in this procedure, HEK293 cells were analyzed by lectin blotting (Supplemental Figure S2), and we observed a notable reduction of at least 50% in the levels of cell-surface glycosylation. Consistent with our previous results, analysis by in-gel fluorescence showed less correlation with UDP-sugar levels (Figure 4). For example, Ac₄GlcNAz, Ac₄GalNAz, and Ac₄GlcNAlk treatment of NIH3T3 cells resulted in similar levels of labeling (Figure 4) despite quite different amounts of UDP-sugar build up (Figure 3). Additionally, the O-GlcNAc-enriched fraction of HeLa cells is more efficiently labeled by Ac₄GlcNAz than Ac₄GalNAz. These data suggest that the selectivity of different MCRs for different classes of glycans could depend both on their ability to be accepted by different glycosyltransferases but also on their metabolic efficiency, which could be different in different cell lines. For example, it is possible that the K_M for the UDP-sugars is lower for OGT than for either the UDP-transporters of the secretory pathway and/or the glycosyltransferases localized in the Golgi.

Figure 4.

Different GlcNAc- and GalNAc-based MCRs label intracellular proteins at varying levels in different mammalian cells. The indicated mammalian cells were treated with one of the four MCRs (@ 200 μM concentration) or DMSO vehicle for 16 h. At this time, the cells were collected by trypsinolysis to enrich the intracellular proteins, lysed, and subjected to CuAAC with either alkynyl- or azido-TAMRA. Labeled proteins were then visualized using in-gel fluorescence scanning. This data is representative of two biological experiments.

In summary, we have examined how the levels of MCR labeling are influenced by the build-up of the corresponding UDP-sugars inside of living cells. Intuitively, we found that the overall levels of MCR incorporation in a specific cell-line correlated very well with the amounts of UDP-sugar. However, this relationship was less clear when different cell-lines were compared to each other. Furthermore, the enrichment of the intracellular fraction of proteins by simple trypsinolysis of the intact cells showed that the labeling of what we presume is mostly O-GlcNAcylated proteins is less dependent on UDP-sugar levels. These data have important implications for the development and application of MCRs. For example, they suggest that efficient labeling of cell-surface glycoproteins may require the use of cell-lines or treatment conditions that result in higher levels of UDP-sugar biosynthesis. In contrast, they also imply that some selectivity for O-GlcNAcylation focused experiments might be achieved by experimental conditions that limit the build-up of the MCR donors. Taken together, we encourage the labs that are involved in the development of MCRs to take these observations under consideration and should complement any cell-based studies with in vitro biochemical analyses when possible. For example, the UDP-sugar of our selective MCR for O-GlcNAcylation, 6AzGlcNAc, has been synthesized and shown directly to be accepted by O-GlcNAc transferase in vitro.^[16] Additionally, the UDP-6AzGalNAc has also been synthesized and shown to be rejected as a substrate by the polypeptide-N-acetyl-galactosamine transferases that initiate mucin O-linked glycosylation,^[13] which directly explains some of its selectivity.