Photoproximity labeling of sialylated glycoproteins (GlycoMap) reveals sialylation-dependent regulation of ion transport

Abstract

Sialylation, the addition of sialic acid to glycans, is a crucial post-translational modification of proteins, contributing to neurodevelopment, oncogenesis, and the immune response. In cancer, sialylation is dramatically upregulated. Yet, the functional biochemical consequences of sialylation remain mysterious. Here we establish a μMap proximity labeling platform that utilizes metabolically inserted azidosialic acid to introduce iridium-based photocatalysts on sialylated cell-surface glycoproteins as a means to profile local microenvironments across the sialylated proteome. In comparative experiments between primary cervical cells and a cancerous cell line (HeLa), we identify key differences in both the global sialome as well as proximal proteins, including solute carrier proteins that regulate metabolite and ion transport. In particular, we show that cell-surface interactions between receptors trafficking ethanolamine and Zinc are sialylation-dependent and impact intracellular metabolite levels. These results establish a μMap method for interrogating proteoglycan function and support a role for sialylated glycoproteins in regulating cell surface transporters.

Graphical Abstract

Introduction

Glycosylation is one of the most common post translational modifications (PTM) on proteins, occurring on at least 50% of all known mammalian proteins and dramatically increasing the functional proteome.¹ Glycosylation can alter both protein localization and function, and misregulated deposition has been shown to contribute to varied disease phenotypes, such as cancer metastasis, viral immune escape, viral entry, and inflammation.^2,3 Glycoproteins also play a critical role in overall cell-surface architecture, contributing to cell adhesion,⁴ cell signaling,⁵ viral docking⁶ and cell-cell interactions.⁷ Among the array of cell-surface monosaccharides, sialic acid stands out as being particularly influential for cell function. This charged sugar is incorporated by sialyltransferases, and commonly decorates the termini of polysaccharide chains (Figure 1). During oncogenesis, overexpression of sialyltransferases leads to hypersialylation,⁸ in turn promoting tumor progression through two different paradigms: (1) Sialylation appears to inhibit apoptosis and allow the cell to evade the immune system,⁹ and (2) The sialoglycoconjugate sialyl Lewis^X facilitates metastasis via extravasation of cancer cells out of the bloodstream into nearby tissue.^10,11

Figure 1.

Utilizing microenvironment mapping to unravel the interactome of sialylated cell-surface glycoproteins.

Despite these observations, the underlying biochemical mechanisms remain poorly understood, in part due to a lack of high-resolution tools to assess the functional roles of sialylation.¹² As such, new methods to profile the biochemical impact of sialylation are needed. We hypothesized that chemoproteomic profiling of both the sialome (sialylated glycoproteins) and associated interactome could provide insights. Protein-protein interactions (PPIs) are typically identified by immunoprecipitation/mass spectrometry (IP/MS) workflows,¹³ whereby the target protein is enriched along with interactors. However, these strategies are challenging for identification of transient or low affinity PPIs, which are especially prevalent in glycoproteins, and often suffer from poor signal to noise.^1,14,15 To surmount these challenges, pioneering work from the Kohler group, and others, has demonstrated that metabolically incorporated sialic acids bearing crosslinkers could be used for photoaffinity labeling (PAL) of sialic acid interactors.^16–23 More recently, Huang and co-workers elegantly showed that galectin-3 fused to APEX2 could be used to interrogate the glycan interactome via peroxide triggered proximity labeling, revealing novel galectin-3 glycoprotein receptors.²⁴ To more comprehensively interrogate sialylation-dependent changes in cancer, we sought to augment our μMap photoproximity labeling platform²⁵ with the capacity to identify the interactome of sialylated cell-surface glycoproteins. This strategy, which we term GlycoMap, employs precise introduction of iridium (Ir) photocatalysts onto sialylated cell-surface glycoproteins using the Bertozzi method²⁶ for metabolic incorporation of azidosialic acid. Thereafter, strain promoted alkyne-azide cycloaddition²⁷ (SPAAC) with a DBCO-tethered iridium catalyst affords iridium-conjugated sialylated glycoproteins. Irradiation with blue light in the presence of a biotin-diazirine probe locally generates carbenes that crosslink with interacting proteins. This strategy offers several distinct advantages in that the introduced proximity labeling machinery is highly selective (iridium incorporation only localized to sialic acid), of relatively small size (as compared to enzymatic proximity labeling strategies), catalytic in nature (allowing for signal amplification), and offers precise spatiotemporal control over labeling.²⁸

Results and Discussion

We began our study by optimizing the iridium catalyst incorporation into the sialome (Figure 2a). Incubation of HeLa cells with tetraacetyl-N-azidoacetylmannosamine (Ac₄ManNAz), followed by treatment with DBCO–iridium (S1) led to incorporation of the iridium photocatalyst onto glycoproteins. Irradiation in the presence of biotin–diazirine (S2) resulted in cell-surface biotinylation as observed by western blot. (Figure 2B). Control reactions displayed minimal biotinylation when omitting the azidosugar, DBCO–iridium reagent, or blue light irradiation. Importantly, immunoprecipitation over streptavidin beads showed strong enrichment of known sialylated glycoproteins nicastrin (NCSTN) and complement decay-accelerating factor (CD55), providing confidence in our workflow (Figure 2C).

Figure 2.

Optimization of a μMap protocol for the photoproximity labeling of sialylated glycoproteins. A) Workflow for the glycomap experiments. Ac4ManNAz incubation performed for 72 h as in (26), however shorter times led to equivalent results in the cell lines in this study. Optimization of Ir-DBCO incubation time is shown in the supporting information (Figure S3). B) Western blot analysis of whole cell lysates after glycomap experiment. C) Western blot of streptavidin enriched lysate following glycomap photoproximity labeling, stained against Nicastrin (top lane) and CD55 (bottom lane). D) Immunofluorescence analysis of cells after glycomap experiment; red: streptavidin, blue: hoechst.

We also compared our method to direct biotinylation of Ac₄ManNAz using DBCO-biotin (S3) and observed labeling of the sialome via western blot. These results suggest our catalytic labeling method can install ~ 0.7 tags per catalyst (see ESI), representing a significant improvement over existing PAL probes, which typically react with water (>95%) and display minimal protein labeling.²⁸ Although the efficiency is slightly lower than direct DBCO-biotinylation, we reason that click-based loading of iridium photocatalysts onto metabolically incorporated Ac₄ManNAz is somewhat more sterically hindered but nonetheless affords photoactivatable biotinylation of glycoproteins and their interactors (Fig 2B). Furthermore, this labeling differential is optimal for distinguishing between glycoproteins and their interactors. Lastly, this protocol was also applicable to both HEK293T cells and primary cervical cells (PCC) (Figure 2D), highlighting the versatility of our workflow. In all cases, confocal microscopy revealed strong biotinylation after treatment with either DBCO–iridium or DBCO-biotin and irradiation in presence of biotin-diazirine. In the absence of azido-sugar no labeling was observed in any case.

Encouraged by these results, we developed a tandem mass tag (TMT)-based quantitative chemoproteomics workflow to identify sialylated cell-surface glycoproteins and map their interactomes (Figure 3A). For each cell type we performed comparative experiments using three different conditions: the first condition (condition A) included SPAAC with DBCO–biotin, which results in biotinylation of only sialylated glycoproteins. Condition B utilized SPAAC with DBCO–iridium for biotinylation of sialylated glycoproteins and their cognate interactome via μMap. Lastly, a control experiment was performed with DBCO–iridium in the absence of Ac₄ManNAz. Together, these parameters allow for the identification of 1) the sialylated glycoproteins, 2) sialylated glycoproteins and their local interactome, and 3) selective identification of protein interactors of the cell surface sialome (GlycoMap). We first examined this chemoproteomic workflow on HEK293T cells (Figure 3B). Using condition A (vs. control) we found significant enrichment (>1.5 Log₂(fold change); >1.5 −Log₁₀(p-value)) of 363 proteins, of which 93% were known glycoproteins, including nicastrin (NCSTN), cadherin 2 (CDH2), small cell adhesion glycoprotein (SMAGP), cluster of differentiation 47 (CD47), basignin (BSG), cluster of differentiation 166 (CD166), cluster of differentiation 99 (CD99) and neuroplastin (NPTN). As predicted by Fig. 3A, condition B (vs. control) enriches both sialylated glycoproteins as well as proximal proteins and thus shares an ~65% overlap with the proteins enriched in condition A (vs. control) (Fig. S9). Analysis of the sialic acid interactome generated by GlycoMap showed 81% of the enriched proteins were membrane-associated, reflecting the accuracy of our labeling approach. In addition, several lysosomal proteins (21) were enriched, presumably arising via internalization of iridium-bound glycoproteins prior to proximity labeling.

Figure 3.

A) Workflow for TMT-based chemoproteomic discovery of interactome of sialylated glycoproteins. Each experiment is performed in triplicate. B) Quantitative chemoproteomics validation by glycomap of HEK293T cells. For all experiments the same cutoffs (>1.5 Log₂(fold change); >1.5 −Log₁₀(p-value)) were used. Selected proteins discussed in the main text are highlighted in gray, red and green. Bottom left: Cryo-EM structure of gamma secretase (PDB code: 5A63), color-coded to the four subunits of the heterotetrameric complex.

In our initial analysis of the GlycoMap dataset we examined a known membrane-bound protein complex, gamma secretase, in order to validate our methodology.

This heterotetramer of membrane proteins (NCSTN, APH1A, PSEN1, PEN-2) proteolytically cleaves many integral membrane proteins, but only NCSTN is directly sialylated.²⁹ In our dataset, NCSTN is highly enriched (3.8 log₂FC) in condition A (vs. control), whereas the non-sialylated interactor, APH1A, is strongly enriched (3.3 log₂FC) in the GlycoMap arm, demonstrating that the μMap workflow can delineate between sialylated glycoproteins and their interactors.

With an established workflow in hand, we next sought to investigate hypersialylation events in oncogenesis. To examine this, we performed comparative GlycoMap experiments on both primary cervical cancer cells (PCC) and the HeLa cervical adenocarcinoma cell line (Figure 4A). In agreement with previous observations of upregulated sialylation, chemoproteomic analysis revealed significantly higher sialylation in HeLa cells (447 enriched proteins) than in PCC (223 enriched proteins) (Figure 4A). This sialome increase in HeLa cells consequently yielded a higher number of interacting proteins (166 enriched proteins in HeLa cells vs 63 enriched proteins in PCC).

Figure 4.

A) Comparative proteomics experiment of primary cervical cells (PCC) and HeLa cells. Top row: sialylated glycoproteins in PCC (left) and HeLa (right). Bottom row: interacting proteins in PCC (left) and HeLa cells (right). Middle: Venn diagram of the enriched proteins from each dataset. The same cutoffs (>1.5 Log₂(fold change); >1.5 −Log₁₀(p-value)) were used for the analysis of all data sets. B) Gene ontology (GO) analysis of the identified sialylated glycoproteins (top) and their interactors (bottom). C) Venn diagram of the enriched solute carrier proteins (SLCs) that interact with sialylated glycoproteins.

We then performed global gene ontology (GO) analysis to categorize the enriched sialylated proteins and their interactors (Figure 4B). Functional enrichment from both cell types was in good agreement with the known roles of identified glycoproteins: cell adhesion, host-cell entry, and regulation of migration, death and defense. Furthermore, examination of GO terms that differed significantly (≥10 Log₁₀ (p-value) change) between the primary and cancerous cervical cells validated that the sialylated glycoproteins in cancerous cells are related to typical oncological phenotypes, including cell morphogenesis, cell-cell adhesion, extracellular matrix organization and tube morphogenesis.

Interestingly, when comparing the roles of identified sialic acid interacting proteins, terms related to small molecule transport were clearly enriched in HeLa cells (organic ion transport, transport of small molecules, vitamin transport), and we specifically noted numerous enriched solute carrier proteins (SLCs) in this dataset (Figure 4C). In particular, intractions of sialylated proteins with SLCs associated with ethanolamine, carnitine, and zinc transport were all significantly enriched in HeLa cells over PCC.

To explore potential consequences of these interactions, we looked for the metabolic consequences of enzymatically depleting them (Figure 5A). Treatment of HeLa cells with sialidase isolated from Vibrio Cholerae (VC-Sia) efficiently cleaves sialic acids α2,3-, α2,6-, or α2,8-linked to cell surface glycans, enabling us to modulate global sialylation status. We incubated HeLa cells either in presence or absence of VC-Sia and then performed mass spectrometry-based metabolomic quantification on cellular metabolite extracts. While most metabolite levels were minimally affected by sialidase treatment, we found that levels of ethanolamine derivatives, including cytidine diphosphate ethanolamine (CDP-Etn) phosphate ethanolamine (P-Etn) and cytidine diphosphate choline (CDP-choline) were significantly increased in sialidase-treated cells (Figure 5B). The solute carrier protein responsible for the transport of ethanolamine, choline-like transporter 1 (SLC44A1),³⁰ is not known to be glycosylated, however our dataset suggests that its function could be regulated by neighboring sialylated glycoproteins. Based on these results, we hypothesize that cell-surface sialic acids may present a negatively charged surface around membrane bound transporters, which in turn could affect the transport of ions, including metabolites.

Figure 5.

A) Workflow for the metabolomics analysis of HeLa cells. B) Metabolite levels of selected small molecules. Experiments were performed in triplicate and C) Left.: GO analysis suggests cation homeostasis is affected by sialylation. Middle: enriched zinc transporters in the HeLa and PCC dataset. Right: Colorimetric zinc assay, which shows a significant change in cellular zinc levels in response to desialylation. P-Values determined by unpaired students t-test. *P < 0 .05, **P < 0.01.

Similarly, we also investigated the impact of sialylation on zinc uptake (Figure 5C). Zinc is imported through the cell membrane by a series of solute carrier proteins of the SLC39 family, four of which are shown in our HeLa dataset to be sialylated (SLC39A6, SLC39A8, SLC39A10 and SLC39A14) and one (SLC39A1) is suggested to interact with a sialylated glycoprotein. Zinc is a key micronutrient that plays a significant role in cell function and its transport is dysregulated in many cancers.^31,32 Using a colorimetric assay to determine the zinc level of untreated and sialidase-treated HeLa cells, we found that the zinc level was significantly higher in cells treated with VC-Sia. These data suggest that cell surface sialylation and/or the interaction with sialylated glycoproteins, plays a role in the regulation of cellular zinc concentration.

Conclusion

In conclusion, hypersialylation in cancer has recently attracted significant interest from the academic and pharmaceutical sectors. However, tools to understand the biochemical consequences of hypersialylation remain limited. Here, we have described a novel proximity labeling platform to identify sialylated cell-surface glycoproteins and their interactors. This sensitive and precise method is robust and compatible with various cell lines, including primary cells. Our comparative proteomics studies between primary and cancerous cervical cell lines show a significant link between sialylation and solute carrier proteins, suggesting a new role for sialylation. Metabolomics data suggest that these interactions regulate the function of certain solute carriers. Together, our platform represents a powerful new approach to sialic acid interactome profiling, providing a systems-level tool for the elucidation of hypersialylation’s biochemical consequences.