Chemical Glycoproteomics

Abstract

Chemical tools have accelerated progress in glycoscience, reducing experimental barriers to studying protein glycosylation, the most widespread and complex form of posttranslational modification. For example, chemical glycoproteomics technologies have enabled the identification of specific glycosylation sites and glycan structures that modulate protein function in a number of biological processes. This field is now entering a stage of logarithmic growth, during which chemical innovations combined with mass spectrometry advances could make it possible to fully characterize the human glycoproteome. In this review, we describe the important role that chemical glycoproteomics methods are playing in such efforts. We summarize developments in four key areas: enrichment of glycoproteins and glycopeptides from complex mixtures, emphasizing methods that exploit unique chemical properties of glycans or introduce unnatural functional groups through metabolic labeling and chemoenzymatic tagging; identification of sites of protein glycosylation; targeted glycoproteomics; and functional glycoproteomics, with a focus on probing interactions between glycoproteins and glycan-binding proteins. Our goal with this survey is to provide a foundation on which continued technological advancements can be made to promote further explorations of protein glycosylation.

Graphical Abstract

1. INTRODUCTION

Mass spectrometry (MS)-based proteomics is a powerful approach for interrogating biological systems. Over the past decade, the in-depth and quantitative characterization of a number of proteomes has advanced our understanding of basic cellular processes. For example, we now have an initial framework for the protein content of discrete cell types and subcellular structures, and we are beginning to define the connectivity and dynamics among those components.^1–6 Accomplishing this has required the acquisition of information-rich data sets that enumerate and quantitate the components of complex protein mixtures. Advances in both MS (i.e., improved mass resolution and accuracy, increased sensitivity and speed, and new fragmentation methods) and computational methods have been critical in this regard.^4,7–11 However, to align MS-based proteomics with the current and future needs for understanding biological processes, proteins also need to be distinguished by their functions, activities, and interactions with other biomolecules.^12–14

Posttranslational modifications (PTMs) are a central mechanism by which cells regulate the functional capabilities of their proteome. Identifying and characterizing these PTMs remains a major goal in proteomics.¹⁵ Chemical approaches have emerged as powerful tools for probing PTMs at the proteome scale.^16–19 Of the more than 300 known protein modifications, glycosylation is one of the most abundant and complex PTMs, and it was one of the first to be probed with chemical tools.^20–22 Indeed, the field of glycobiology has long been driven by innovations from chemistry, such as glycan and lectin microarrays, small-molecule inhibitors, and chemically synthesized glycans.^21,23,24

More recently, chemists have focused on developing tools to help profile protein glycosylation at the level of the proteome. In this review, we discuss key contributions from chemical biology for interrogating the “glycoproteome” by MS. We begin with a discussion of the importance of and challenges in studying protein glycosylation. We then survey chemical approaches for the enrichment of glycoproteins and glycopeptides from complex mixtures, emphasizing methods that exploit the unique chemical properties of glycans or introduce unnatural functional groups through metabolic labeling and chemoenzymatic tagging. Next, we summarize strategies for identifying sites of protein glycosylation and approaches for targeted glycoproteomics. Finally, we conclude with a discussion of functional glycoproteomics and how chemical tools can be used to detect glycoprotein interactions. The literature covered in this review spans a time period from the 1970s through 2016, with a majority of the work occurring from 2000 onward.

2. GLYCOSYLATION IS AN IMPORTANT PTM THAT INFLUENCES PROTEIN ACTIVITY AND FUNCTION

Protein glycosylation mediates a diversity of cellular processes. It is predicted that 20% of all eukaryotic proteins are glycosylated,²⁵ but this is likely an underestimation. Unlike DNA transcription and RNA translation, glycosylation is not template-driven; instead, glycoproteins are generated through the enzymatic addition of monosaccharides and complex oligosaccharides onto protein scaffolds, a process that can produce significant diversity and heterogeneity.^26,27 The attached glycans can endow the protein with new properties including cellular trafficking and localization, binding specificity, and thermodynamic stability.^28–30 Glycoproteins on the cell surface facilitate cell recognition, signaling, and adhesion.^31–33 Moreover, changes in protein glycosylation—including in the glycan structure and site occupancy—can modulate cellular phenotypes such as growth, development, and disease.^34–36 Importantly, aberrant glycosylation is well recognized as a hallmark of many diseases and their progression, including microbial pathogenesis, immune deficiencies, neurodegenerative diseases, and many cancers.^37–39 This phenomenon has been exploited through the development of more than 15 glycoproteins as clinical biomarkers.^38,40–43 Consequently, a large effort has been directed toward the accurate characterization of a cell’s glycoproteome.

Characterizing the glycoproteome is a challenging endeavor. The nine monosaccharide building blocks used in vertebrates can be assembled in diverse linear and branched patterns to generate a staggeringly complex collection of carbohydrates in an organism, also known as its “glycome.” Additional structural diversity results from the attachment of monosaccharides to specific sites on protein scaffolds.^26,44,45 These modifications produce different classes of protein glycosylation, including N-linked, O-linked, and C-linked, with the first two being the most common and the focus of this review.^46–49 In N-glycosylation, glycans are attached to the side-chain nitrogen atoms of asparagine (Asn, N) residues found primarily within the consensus sequence N–X–S/T (where X is any amino acid except proline).⁵⁰ N-glycosylation is frequently found on proteins that are translated into the endoplasmic reticulum (ER), where a universal glycan core structure, beginning with a β-linked N-acetylglucosamine (GlcNAc), is transferred onto nascent proteins (Figure 1a).⁵¹ The N-glycan precursors are enzymatically edited to yield a plethora of mature glycan structures with compositions that largely depend on cell type, protein localization, and protein structure.⁵² However, a majority of N-glycans share a conserved pentasaccharide core structure even after maturation (Figure 1a, red highlight). This feature enables proteome-wide manipulation of N-glycoproteins. As described later, several methods for identifying N-glycoproteins capitalize on the universality of the conserved core structure for lectin-based enrichment or glycan cleavage using specific enzymes.

Figure 1.

Glycan structures display a range of complexity. (a) Example of an N-glycan, with the conserved pentasaccharide (Man₃GlcNAc₂) core structure highlighted in red. This core structure can be further elaborated with N-acetyllactosamine (GalGlcNAc, green), sialic acid (purple), or other monosaccharides. (b) Example of a mucin-type O-glycan, with the conserved α-linked GalNAc residue highlighted in red. This monosaccharide can also be elaborated to generate higher-order glycans; depicted is the sialyl Lewis X antigen (blue). (c) Example of the O-GlcNAc modification, characterized by a single β-linked GlcNAc monosaccharide.

In contrast to N-glycosylation, O-glycosylation in vertebrates encompasses considerable structural diversity without a universal core structure. O-glycans are attached to the side-chain oxygen atom of serine (Ser), threonine (Thr), tyrosine, or hydroxylysine residues.^47,53–55 Each type of O-glycosylation is defined by the identity of the peptide-proximal glycan structure. The most common is the mucin-type, which is initiated by an α-linked N-acetylgalactosamine (GalNAc) residue attached to Ser or Thr residues (Figure 1b).²⁶ O-glycans can also be initiated with Ser- or Thr-bound glucose, xylose, mannose, or fucose residues.²⁶ These core monosaccharides are then elaborated by a series of glycosyltransferases to generate high-order structures, such as the tetrasaccharide sialyl Lewis X (Figure 1b, highlighted in blue), and can be found on secreted or cell-surface glycoproteins. One unique subclass of O-glycosylation, found in the cytoplasm and nucleus of most eukaryotic cells, is the addition of a single β-linked GlcNAc residue onto either Ser or Thr (Figure 1c) by the enzyme O-GlcNAc transferase (OGT).^56–58 Termed O-GlcNAcylation, this intracellular form of glycosylation is a rapidly reversible, substoichiometric modification, analogous to phosphorylation.⁵⁹ The modification is involved in a number of signaling pathways, including cellular metabolism and physiology.⁶⁰

The complexity and heterogeneity inherent to protein glycosylation impart an analytical and computational challenge to the study of this class of PTM. Although a number of biochemical techniques are available to obtain crude structural information regarding a protein-bound glycan, including the use of glycan-binding proteins (i.e., lectins), radioactive labeling, antibodies, and enzymatic deglycosylation, they all suffer from low information content and cumbersome manual analysis. Analytical methods such as NMR spectroscopy, liquid chromatography, and electrophoresis can also be used to characterize protein-associated glycans, but their utility is limited by sample complexity and, in many cases, by the quantity of sample available.⁶¹

As an alternative to these methods, liquid chromatography coupled to tandem mass spectrometry (LC-MS/MS) has emerged as a powerful technology for profiling protein glycosylation. A number of excellent reviews on the subject of MS analysis of glycoproteins have been published.^62–72 The methods primarily employ bottom-up shotgun proteomics, a technique illustrated in Figure 2.^7,9,73,74 In this method, proteins are digested to produce a mixture of peptides, which are then separated by LC and analyzed by MS. Peptide ions in the full-scan mass spectrum are selected for tandem MS analysis, typically on the basis of ion intensity, to generate fragment-ion mass spectra. Comparison to a proteome database allows the assignment of the tandem mass spectrum to its most likely peptide sequence and, ultimately, parent protein of origin.^11,75,76 Despite successes, glycoproteomics remains challenging for a number of reasons. Glycosylation is a labile PTM and requires technical expertise to ensure that the glycan is preserved in its entirety during sample handling or during MS analysis.^77,78 Many interesting glycoproteins are typically present in low abundance, increasing the difficulty of detection in complex biological samples such as human plasma, where protein concentrations can span up to 12 orders of magnitude.^79,80 Adding to the challenge, glycoproteins usually exist as complex mixtures of glycosylated variants called glycoforms. These structures exhibit both microheterogeneity (i.e., different glycan structures at a specific site of modification) and macroheterogeneity (i.e., differences in the location and number of glycan modifications of the glycoprotein). Additional complexity is introduced through the linkage stereochemistry between monosaccharides and because several monosaccharides are structural isomers of each other with the same molecular weight. These variations complicate the identification process and pose problems for standard bioinformatics software that cannot cope with the enormous variability in glycoprotein structures for a given molecular weight.^81–83

Figure 2.

Schematic of the bottom-up shotgun proteomics method. (a) A mixture of proteins is subjected to proteolytic digestion, producing (b) a mixture of peptides. (c) The peptides are separated using LC, and (d) full-scan mass spectra are collected at regular intervals. (e) Typically, ions are selected from each full scan for fragmentation (i.e., tandem MS) in a data-dependent fashion (i.e., based on ion abundance). (f) These tandem mass spectra are analyzed using a database search to generate peptide and protein identifications.

Considering these formidable obstacles, the successful identification and characterization of glycoproteins by MS will require a combination of new methods, from sample preparation to data analysis.⁸⁴ In particular, enrichment through the reduction of sample complexity has played a significant role in the study of glycoproteins in complex mixtures. Following sample preparation, glycoproteins can be characterized using three complementary LC-MS methods: glycoproteomics, proteomics, and glycomics. Although these terms are sometimes used interchangeably and/or inconsistently, glycobiologists have started to adopt specific definitions.^85–87 Here, we define glycoproteomics as the analysis of intact glycopeptides to simultaneously obtain information about the peptide, the site of modification, and the glycan; glycomics as the analysis of free or released glycans from glycoproteins; and proteomics as the analysis of deglycosylated peptides or the nonglycosyated glycoprotein peptides. In this review, we highlight the role of chemical tools in both glycoproteomic and proteomic analyses of glycoproteins; strategies for glycan analysis have been summarized in a number of glycomics reviews.^{28,69,88–91}

3. ENRICHMENT OF GLYCOPROTEINS AND GLYCOPEPTIDES

3.1. Overview

Reducing interference from more abundant and nonglycosylated species is an important step for the in-depth analysis of glycoproteins in complex mixtures using MS. Typically, enrichment has been achieved by affinity purification using antibodies or lectins that target specific classes of glycoproteins.^38,92,93 Lectins are glycan-binding proteins isolated from various plants, bacteria, and animal sources and have played an important role in the isolation, purification, and characterization of glycoproteins.^94,95 This is most prevalent for glycoproteins bearing N-glycans, where lectins with varying glycan-binding specificities have been used for enrichment.⁶² For example, Concanavalin A (ConA), which has been used extensively to enrich glycoproteins, has a strong preference for oligomannose-type N-glycans compared to highly branched or complex-type N-glycans.⁶² Lectins that recognize peripheral glycan structures have also been used for enrichment, such as Sambucus nigra agglutinin (SNA) and Aleuria aurantia lectin (AAL), which have varying affinities to sialic acid and fucose residues, respectively; the lectins’ binding promiscuity is based on the monosaccharide’s linkage pattern.⁹⁶ Nevertheless, these lectins were immobilized on solid supports and used to enrich glycoproteins from human plasma for biomarker detection.⁹⁷ In a similar manner, a combination of lectins were placed in an ultrafiltration spin filter [i.e., filter-aided sample preparation (FASP)] for enrichment and used to identify over 2000 N-glycoproteins from different mouse tissues.⁹⁸ This method, termed N-glyco FASP, was also used to profile N-glycosites across seven evolutionarily distant species⁹⁹ and was combined with stable isotope labeling by amino acids in culture (SILAC) to differentiate between different patient-derived B-cell lymphoma cell lines.¹⁰⁰

In contrast to N-glycans, the greater diversity of O-glycan structures restricts the utility of any one lectin as a universal capture agent. In one strategy, serial lectin affinity chromatography, N-glycopeptides were first removed using ConA and then GalNAc-containing peptides were enriched using the lectin Jacalin.¹⁰¹ For O-GlcNAcylation, the most common lectin enrichment method uses wheat germ agglutinin (WGA),¹⁰² which has also been used in a long-column format, termed lectin weak-affinity chromatography (LWAC), to enrich for O-GlcNAcylated peptides.^103,104 Recently, Solanum tuberosum lectin was discovered to recognize O-GlcNAc and then used to enrich glycoproteins from colon cancer tissue.¹⁰⁵ In addition to lectins, several antibodies have also been used to enrich O-GlcNAcylated glycoproteins from complex mixtures.^106–108 Unfortunately, the utility of lectins and antibodies for glycoprotein and glycopeptide enrichment is limited by low substrate binding affinities and/or poor specificities.

Toward addressing these limitations, advancements in chromatographic and solid-phase extraction (SPE) techniques have been applied to glycopeptide enrichment (see these recent reviews for more information).^{66,68,84,86,87,109,110} One method is based on hydrophilic interaction liquid chromatography (HILIC), which has enabled a number of glycoproteomics studies by providing a general method for separating and enriching glycopeptides from nonglycosylated peptides.^111,112 Improvements to HILIC for glycopeptides have taken shape in many formats, including the exploration of different stationary phases, introduction of ion-pairing reagents in the mobile phase, and implementation in an SPE format.^112–114 Other strategies include the use of titanium dioxide-functionalized materials to enrich for sialylated glycopeptides¹¹⁵ or boronic acid-functionalized materials as a more general enrichment method (see section 3.2 for more information).¹¹⁶

There is also a growing interest in the use of chemical tools for marking and enriching subproteomes for analysis by MS.^117,118 In early work, chemical probes were developed to target specific amino acids for quantitative proteomics.¹¹⁷ The progenitors of this approach were the isotope-coded affinity tag (ICAT) and isobaric tags for relative and absolute quantitation (iTRAQ), which were used to label cysteine residues or reactive amines, respectively, for quantitative, comparative proteomics.^119,120 Subsequently, chemical probes that targeted specific classes of enzymes were developed to enrich functionally active species of the proteome. Termed activity-based protein profiling (ABPP), this method employs families of enzymes that are covalently modified with small molecule probes that contain a chemical warhead specific for their active site, allowing for the identification and characterization of many classes of enzymes, including proteases, glycosidases, kinases, and phosphatases.^121–124

Chemical glycoproteomics methods take advantage of glycans to mark, enrich, and identify glycoproteins or glycopeptides by MS. For example, glycoproteins and glycopeptides can be specifically labeled at their site(s) of glycosylation with a chemical reporter.¹²⁵ The reporter can serve as a handle for either direct analysis or enrichment after covalent modification with a secondary probe. The selections of the chemical reporter and secondary probe are important to ensure that the reaction between the two functional groups is specific and efficient and can take place in a biological environment; such characteristics define a “bioorthogonal” chemical reaction (Figure 3).^126–131 There are three main strategies for introducing a bioorthogonal chemical reporter at sites of glycosylation. The first relies on exploiting chemical reactivity unique to the structure of a monosaccharide (for example, the cis diol in galactose or sialic acid). The second method relies on metabolic labeling, wherein an unnatural monosaccharide is incorporated directly into glycans using the cell’s biosynthetic machinery. The third approach relies on using purified enzymes to chemoselectively append a chemical reporter onto a specific monosaccharide in vitro. Each of these three strategies is discussed in turn as it relates to the enrichment of glycoproteins for MS.

Figure 3.

Bioorthogonal reactions with aldehyde-, ketone-, azide-, or alkyne-functionalized monosaccharides. These chemical handles can be introduced into glycans through chemical, metabolic, or enzymatic methods. After an enrichment probe (red star) has been appended to glycoproteins or glycopeptides, these compounds can be enriched from complex mixtures and identified by MS. (a) Aldehydes and ketones are condensed with amine nucleophiles, including hydrazide compounds (top) and aminooxy compounds (bottom), to form stable hydrazone or oxime linkages, respectively. (b) Azides undergo the Staudinger ligation (top) with triarylphosphines or a [3 + 2] cycloaddition with terminal alkynes through the Cu-catalyzed azide–alkyne cycloaddition (CuAAC, middle) or strained alkynes in the absence of Cu (copper-free click chemistry, bottom). (c) Alkynes undergo a [3 + 2] cycloaddition with azides through the CuAAC reaction.

3.2. Enrichment by Exploiting Chemical Properties of Carbohydrates

An early approach to enriching glycoproteins from complex mixtures utilized the cis-diol functionality present in many monosaccharides (for example, see Figure 4a, where a cis diol is highlighted in gray). Cis diols can form reversible covalent adducts with boronic acids, or they can be converted to aldehydes after oxidative cleavage, which, in turn, form covalent adducts with hydrazide or aminooxy reagents. These chemistries allow selective conjugation to solid-supported matrixes or affinity probes, enabling enrichment of the glycosylated species.

Figure 4.

Enrichment of glycoproteins using carbonyl chemistry. (a) Schematic of hydrazide capture technology. (1) Glycoproteins are subjected to mild oxidants, most commonly sodium periodate, which oxidatively cleaves the cis diols (highlighted in a gray box) in glycans into an aldehyde group (highlighted in purple). (2) The aldehyde is condensed with a hydrazide-functionalized solid support to covalently capture and enrich glycoproteins. (3) The protein is released by deglycosylation from the solid support for identification by LC-MS. (b,c) Hydrazide capture technology can also target specific monosaccharides, including (b) sialic acid using optimized oxidation conditions and (c) galactose using galactose oxidase.

3.2.1. Boronic Acid Adduct Formation with Cis Diols

Boronic acid-functionalized materials have long been used as chromatography supports for the separation of glycans.^116,132,133 When boronic acids are attached to a solid support, such as agarose particles, magnetic beads, or a mesoporous silica matrix, boronic esters are formed with glycans under basic conditions, enriching both O- and N-glycoproteins.^134,135 Under acidic conditions, glycoproteins are released and can be analyzed and identified by tandem MS. This approach has been employed to detect glycoproteins in human blood samples¹³⁶ and N-glycoproteins in human saliva¹³⁷ and yeast.¹³⁸ To improve the specificity, boronic acid-decorated lectins (BAD-lectins) have been used to enrich specific glycans or glycoproteins from complex mixtures.¹³⁹

3.2.2. Periodate Oxidation to Introduce an Aldehyde

Another enrichment strategy that exploits the chemical properties of glycans is the introduction of aldehyde groups by sodium periodate-mediated oxidative cleavage of cis diols. The aldehyde group serves as a chemical reporter and can react with amine nucleophiles, including hydrazine and aminooxy compounds, to form covalent linkages under physiological conditions (Figure 3a). Aebersold and co-workers were the first to implement this strategy in a glycoprotein profiling experiment: They used hydrazide-functionalized beads to enrich periodate-oxidized glycoproteins from human serum (Figure 4a).¹⁴⁰ With a covalent bond between the glycoproteins and the solid support, nonglycosylated proteins were washed away, and the nonglycosylated portions of the bound glycoproteins were digested away using trypsin. The enzyme peptide:N-glycosidase F (PNGase F), which selectively cleaves the amide bond linking N-glycans to asparagine side chains, was used to release the covalently bound N-glycopeptides from the support, which were then identified by tandem MS. Using this “hydrazide capture technology”, the authors identified 145 N-glycosites, mapping to 57 N-glycoproteins, in human serum samples. Subsequently, Zhang and co-workers developed a variant of this method in which glycoproteins are first digested into peptides and then treated with sodium periodate.¹⁴¹ A number of studies have applied either one of these methods for N-glycosite mapping in human plasma, cerebrospinal fluid, islets, platelets, saliva, the microsomal fraction of an ovarian cancer cell line, and mouse embryonic stem cells.^42,142–150 Most recently, Zhang and co-workers used hydrazide capture technology to perform an integrated proteomic and glycoproteomic analysis of two prostate cancer cell lines, LNCaP and PC3. By analyzing both peptides and intact glycopeptides, the authors were able to compare differences in protein abundance and glycan composition at specific glycosites between the cell lines.¹⁵¹ Finally, Zou and co-workers developed variants of hydrazide capture technology using hydroxylamine to assist in releasing captured glycopeptides¹⁵² or reductive amination to block N-terminal amines on glycopeptides prior to oxidation of glycans.¹⁵³

There are, however, some limitations to hydrazide capture technology. First, hydrazide-functionalized beads can react with naturally occurring carbonyl-containing biomolecules in cell lysates (including glucose, pyruvate, and pyridoxal-5′-phosphate), which will reduce the selectivity and sensitivity for glycoprotein enrichment. Second, there is a lack of specificity in both the monosaccharides that are oxidized and the class of glycoproteins that are enriched (i.e., both O- and N-glycoproteins can contain oxidized glycans). Whereas the use of PNGase F to release captured glycopeptides will bias the analysis to N-glycosylation sites, a similar strategy is not available for O-glycosite profiling. Finally, side reactions can occur during glycan oxidation with sodium periodate (i.e., oxidation of N-terminal Ser/Thr, cysteine, and methionine residues).¹⁵⁴ Despite these limitations, it should be noted that hydrazide capture technology has broad applicability to a variety of biological samples, including human biofluids.¹⁴⁴

To improve the sensitivity of hydrazide capture technology for enriching cell-surface glycoproteins, Wollscheid et al. optimized sodium periodate oxidation conditions for live-cell labeling.¹⁵⁵ This strategy enabled treatment of the newly generated aldehydes with a hydrazide-functionalized enrichment probe (biocytin hydrazide) without lysing cells and exposing the intracellular and naturally occurring carbonyl-containing biomolecules. Following proteolytic digestion of labeled-cell lysates, glycopeptides from cell-surface glycoproteins were captured with streptavidin beads. After enzymatic release using PNGase F, the formerly N-glycosylated peptides were identified by tandem MS. This method, termed cell-surface capturing (CSC) technology, was used to identify the differences between the cell-surface N-glycoproteome of T and B cells. In the same work, CSC technology was combined with SILAC to monitor the N-glycoproteome dynamics during T-cell activation and during mouse embryonic-stem-cell differentiation into neurons. CSC technology has also been applied to mouse myoblast cells, mouse β-cells, hepatoma cells, leukemia cells, pluripotent stem cells, Drosophila Kc167 cells, and neuronal progenitor cells.^{145,156–163} Recently, Wollscheid and co-workers used CSC technology to identify cell-surface proteins in 41 human and 31 mouse cell lines.¹⁶⁴ This work is summarized in a publically available resource: the Cell Surface Protein Atlas (CSPA).

Although CSC technology reduces cross-reactivity between the enrichment probe and native carbonyl-containing bio-molecules, the use of PNGase F restricts the method to N-glycoprotein identification. Larson and co-workers addressed the limitation by optimizing sodium periodate oxidation conditions to selectively oxidize the cis diol in sialic acid residues, a terminal monosaccharide found on both N- and O-glycoproteins (Figure 4b).¹⁶⁵ After oxidized glycoproteins were captured on hydrazide beads and nonglycosylated peptides were digested off using trypsin, both N- and O-glycopeptides were released using mildly acidic conditions that cleave sialic acid residues from the underlying glycan structure. When applied to human cerebrospinal fluid, the authors identified 36 N-glycopeptides and 44 O-glycopeptides, mapping to a total of 45 glycoproteins. Using a nested fragmentation method (i.e., MS³) and manual annotation, the authors could distinguish between N- and O-glycopeptides and characterized some of the intact glycopeptide structures. This strategy has also been used to identify glycoproteins in human urine and mouse serum.^166,167

In a similar fashion, Paulson and co-workers demonstrated that it is possible to selectively introduce aldehydes into cell-surface sialic acid residues using mild sodium periodate oxidation conditions.¹⁶⁸ They improved the efficiency of chemical labeling through analine-catalyzed oxime formation with an aminooxy-biotin enrichment probe. Following enrichment with streptavidin beads, the nonglycosylated peptides were digested off using trypsin and used to identify putative glycoproteins, whereas N-glycopeptides were cleaved off using PNGase F and used to identify N-glycosites.¹⁶⁹

Recently, hydrazide capture technology was extended to the O-GlcNAc modification. Medzihradszky and co-workers demonstrated that the trans diol of GlcNAc can be oxidatively cleaved into aldehydes using high concentrations of sodium periodate and elevated temperatures.¹⁷⁰ After oxidized glycoproteins were captured on a hydrazide-functionalized resin, nonglycosylated peptides were removed using trypsin, and glycopeptides were obtained by treatment with hydroxylamine under mildly acidic conditions. The authors applied this method to study the Drosophila melanogaster proteasome and identified 12 O-GlcNAcylated peptides corresponding to 5 glycoproteins.

In summary, hydrazide capture technology and its derivatives have become valuable chemical enrichment strategies for identifying N-glycoproteins. These technologies have enabled the identification of thousands of new glycoproteins across many organisms. Although some effort has been made to extend the method to O-glycoproteins, the lack of efficient chemical and enzymatic tools for release has limited its utility in this respect.

3.2.3. Enzymatic Oxidation to Introduce an Aldehyde

Hattori and co-workers demonstrated that hydrazide capture technology could be used to enrich two proteins bearing two forms of O-glycosylation unique to collagen: galactosyl hydroxylysine (GHL) and glucosyl galactosyl hydroxylysine (GGHL).¹⁷¹ Instead of chemically oxidizing the glycans with sodium periodate, which was previously shown to degrade these structures, they used galactose oxidase, an enzyme that oxidizes terminal galactose and GalNAc residues at the C-6 position to generate an aldehyde (Figure 4c). The authors applied the technique to enrich glycopeptides from bovine type I and type II collagens. After enriching oxidized glycopeptides on hydrazide resin and eluting them under acidic conditions, they identified 37 GHL/GGHL glycopeptides using tandem MS, 13 from type I and 24 from type II collagen. Recently, Paulson and co-workers extended the utility of galactose oxidase to live cells and demonstrated that aldehydes could be selectively introduced into terminal galactose and GalNAc residues on cell surfaces.¹⁶⁹ Toward expanding the substrate scope of galactose oxidase, directed evolution was used to select for enzyme variants that could oxidize mannose and GlcNAc residues.¹⁷² Recently, Kohler and co-workers utilized both enzymatic and chemical oxidation to enrich and identify mammalian glycoprotein substrates for bacterial sialidases.¹⁷³

3.3. Enrichment by Metabolic Labeling with Unnatural Monosaccharide Analogues

Metabolic labeling with chemically functionalized monosaccharide substrates is an emerging strategy for glycoprotein profiling.^{16–18,63,70,174–177} The approach can be traced to classic work from the 1970s indicating that glycan biosynthetic enzymes are permissive, in some cases, of unnatural substrates, enabling their metabolic incorporation into glycoproteins. In the 1990s, Reutter and co-workers reported that the N-acyl group of sialic acid can be biosynthetically modulated by feeding cells and animals N-acyl derivatives of its metabolic precursor, N-acetylmannosamine (ManNAc).^178,179 These observations suggested a route for incorporating bioorthogonal functional groups into glycoproteins as a means of visualizing them in cells and organisms.^180–184 In the past decade, interest has grown in using metabolic labeling to tag and enrich specific classes of glycoproteins for proteomic analysis. As shown in Figure 5a, cells or organisms are treated with a modified sugar, and the metabolically labeled glycoprotein products are chemically tagged with an enrichment probe. The bioorthogonal chemistries employed for such studies generally involve the azide as a principle reagent. Azides are small in size, easily added to synthetic sugar analogues, stable in vivo, and unreactive with normal cellular components.^126,130 After metabolic incorporation into glycoproteins, azido sugars can be selectively reacted by the Staudinger ligation with triarylphosphines,¹⁸⁰ copper(I)-catalyzed azide–alkyne cyclo-addition (CuAAC) with terminal alkynes,^185,186 or “copper-free click chemistry” with strained cyclooctynes (Figure 3b,c).^187,188 Alternatively, simple sugars can be functionalized with terminal alkynes, which are similar in size to azides, and their metabolic products labeled with azide probes.^126,130

Figure 5.

Enrichment of glycoproteins using metabolic labeling followed by bioorthogonal ligation. (a) Schematic of metabolic labeling to introduce an unnatural monosaccharide carrying a chemical reporter (i.e., azide or alkyne; purple triangle) directly into glycans. (1) Unnatural monosaccharides (blue hexagons) are accepted by the cell’s natural machinery and incorporated into the cell surface, secreted, and/or attached to intracellular glycoproteins (orange ovals). The natural monosaccharides are depicted as gray hexagons. (2) A bioorthogonal chemical reaction is performed to label glycoproteins with a probe molecule (red star). (b) Chemical structures of natural monosaccharides (left column) and their azide and alkyne analogues (functional groups highlighted in purple) that have been used for metabolic labeling and enrichment of glycoproteins or glycopeptides. Although the unnatural monosaccharides are depicted in a peracetylated form, after they have entered cells by passive diffusion, cytosolic esterases will cleave the acetyl groups.

Metabolic labeling can be used to target specific forms of protein glycosylation for enrichment. For example, sialylated glycoproteins can be selectively labeled using analogues of its biosynthetic precursor, ManNAc. Toward this end, both azido (ManNAz) and alkynyl (ManNAlk) derivatives of ManNAc have been employed (Figure 5b).^180,183,189 Alternatively, sialic acid analogues can be used directly as metabolic labels, circumventing the requirement that upstream enzymes recognize unnatural substrates.^190,191 Likewise, azido and alkynyl variants of GalNAc,^192–194 fucose,^189,195,196 and GlcNAc^197–199 have been used to label and enrich various glycoprotein types. These sugars are predominantly administered in a per-O-acetylated form to improve membrane permeability and efficient cell entry. After the sugars have entered the cells, cytosolic esterases cleave the acetyl groups, and glycan biosynthetic enzymes can utilize the deprotected sugar derivatives.²⁰⁰

3.3.1. Labeling Sialylated Glycoproteins with ManNAc Analogues

There has been much interest in the selective profiling of sialylated glycoproteins due to known associations of changes in cellular sialylation states during tumor progression.^201,202 To target this group of glycoproteins, both per-O-acetylated N-azidoacetylmannosamine (Ac₄ManNAz) and per-O-acetylated N-alkynylacetylmannosamine (Ac₄ManNAlk) have been used to metabolically label cellular glycans (Figure 5b).

In an early example, Wong and co-workers fed prostate cancer-derived PC3 cells Ac₄ManNAlk and then treated whole-cell lysates with a biotin-azide probe under CuAAC reaction conditions.²⁰³ After enrichment of sialic acid-containing glycoproteins using streptavidin beads, nonglycosylated peptides were removed by on-bead tryptic digestion, and then N-glycopeptides were released using PNGase F. Following tandem MS analysis of the released peptides, the authors identified 219 N-glycosites, corresponding to 108 glycoproteins, of which 69 were previously known glycoproteins.

Semmes and co-workers also combined metabolic labeling and MS to identify sialic acid-containing glycoproteins involved in prostate cancer.²⁰⁴ The authors treated highly metastatic (ML2 cells) and nonmetastatic (N2 cells) PC3-derived cell lines with Ac₄ManNAz and performed the CuAAC reaction on intact cells with a biotin-alkyne tag. After whole-cell lysates had been prepared and biotin-labeled glycoproteins had been captured onto streptavidin beads, glycoproteins were released by boiling and then subjected to in-gel digestion with trypsin. Following tandem MS, the authors identified 324 proteins unique to N2 cells and 372 proteins unique to ML2 cells, of which 36 and 44 proteins, respectively, were predicted to be cell-surface N-glycoproteins.

Another example of metabolic labeling with Ac₄ManNAz was performed by Bai and co-workers.²⁰⁵ The authors fed three different cancer cell lines (A549 lung adenocarcinoma, HeLa cervical carcinoma, and SW1990 pancreatic adenocarcinoma) Ac₄ManNAz, performed the CuAAC reaction on intact cells using a thiol-cleavable biotin-alkyne probe, prepared whole-cell extracts, and captured the biotinylated proteins on magnetic streptavidin beads. After chemical release with dithiothreitol, proteins were digested using trypsin and analyzed by tandem MS. Collectively, 310 proteins were identified, 56 of which were predicted to be cell-surface glycoproteins.

In addition to cell-surface glycoproteins, metabolic labeling with Ac₄ManNAz was used to identify glycoproteins in the secretome.²⁰⁶ Drake and co-workers cultured two stromal cell lines (WPMY-1 and HS5) in serum-free or low-serum media containing Ac₄ManNAz. The clarified cell culture media were treated with alkyne beads to capture glycoproteins. Using on-bead tryptic digestion and tandem MS, the authors identified over 100 secreted proteins in each stromal cell line’s secretome, many of which were previously known glycoproteins. In the same study, the authors also investigated the utility of metabolic labeling with per-O-acetylated N-azidoacetylgalactosamine (Ac₄GalNAz) for identifying glycoproteins in the secretome.

To study glycoprotein dynamics in vivo, Chen and co-workers used Ac₄ManNAz and the CuAAC reaction to identify glycoproteins upregulated during cardiac hypertrophy in live rats.²⁰⁷ After administration of Ac₄ManNAz to saline- or isoprenaline-treated rats, heart tissue was isolated, lysed, and treated with an alkyne-biotin probe. Following enrichment of biotinylated glycoproteins with streptavidin beads and in-gel digestion, the authors used tandem MS to identify newly synthesized sialylated glycoproteins in hypertrophic hearts. They also demonstrated that Ac₄GalNAz could be used to label heart glycoproteins.

Copper-free click chemistry has also been used to enrich glycoproteins after metabolic labeling with ManNAc analogues. In one example, Lichtenthaler and co-workers used a cyclooctyne probe to enrich for glycoprotein substrates of membrane-bound proteases (i.e., sheddases).²⁰⁸ The authors enriched glycoproteins from the secretome of primary neurons by culturing cells in Ac₄ManNAz, treating the clarified media with a biotin-dibenzocyclooctyne probe, and then incubating with streptavidin beads. Following release and tryptic digestion, they identified 34 substrates for the protease BACE1, an Alzheimer’s-relevant sheddase.

In another example, Tackett and co-workers used metabolic labeling to enrich and identify cell-surface glycoproteins from pancreatic cancer cell lines.²⁰⁹ The authors cultured three pancreatic cell lines (BxPC-3, MIAPaCa-2, and Panc-1) in Ac₄ManNAz and then treated the cells with a biotinylated and thiol-cleavable cyclooctyne probe. Cell lysates were prepared and incubated with streptavidin beads to capture glycoproteins. Following cleavage with a thiol reagent, the authors used tandem MS to identify several hundred proteins between the three cell lines.

In addition to the CuAAC reaction and copper-free click chemistry, the Staudinger ligation has also been combined with metabolic labeling with ManNAc analogues to enrich for glycoproteins. In one study, Autelitano et al. identified biomarkers for malignant glioma targeted therapy.²¹⁰ Primary cells isolated from human glioblastoma multiform (GBM) tissues and nontumorous brain tissue were cultured in Ac₄ManNAz and treated with a biotinylated phosphine probe to label cell-surface glycoproteins. Biotinylated glycoproteins were enriched from cell lysates using streptavidin beads and eluted through on-bead digestion with trypsin. Using tandem MS, the authors identified 801 proteins, 606 of which were predicted to be glycoproteins. Additionally, using label-free quantification, the authors identified 35 overexpressed proteins in GBM tumors and validated their expression profiles using gene expression analysis.

In another study, Delcourt et al. identified sialylated glycoproteins in hypoxic human umbilical vein endothelial cells.²¹¹ Cells were cultured in Ac₄ManNAz under hypoxic conditions and then treated with a biotinylated phosphine probe to label cell-surface glycoproteins. Biotinylated glycoproteins were enriched from cell lysates using streptavidin beads. Following on-bead digestion and tandem MS, the authors identified 27 proteins upregulated in the hypoxia-treated cells, many of which were previously known sialylated glycoproteins or were validated by Western blot analysis.

Most recently, Zhang and co-workers identified N-glycoproteins as potential biomarkers for pancreatic cancer.²¹² The authors cultured SW1990 human pancreatic cells in Ac₄ManNAz and biotinylated azido-glycoproteins using the Staudinger ligation. After biotinylated glycopeptides had been captured using streptavidin beads, the formerly N-glycopeptides were selectively released using PNGase F and analyzed by tandem MS. The authors identified 75 N-glycosites across 55 glycoproteins, 42 of which were linked to cancer. Using immunohistochemistry, they proposed two new pancreatic cancer biomarkers: lysosome-associated membrane protein 1 and hypoxia upregulated protein 1.

3.3.2. Labeling Sialylated Glycoproteins with Sialic Acid Analogues

A major challenge with metabolic labeling in whole organisms is achieving tissue-specific incorporation of unnatural monosaccharide analogues. Chen and co-workers overcame this limitation by encapsulating monosaccharides in ligand-targeted liposomes to direct sugar analogues to tumor cells in vivo.^213,214 The authors encapsulated 9-azido sialic acid into a liposome displaying a cyclic-RGD peptide, injected the construct into a melanoma xenograft mouse model, and observed high expression of azido sialic acid in tumor cells. The authors collected tumor tissue samples at different time points, treated tissue lysates with an alkyne-biotin probe, and enriched for glycosylated proteins using streptavidin beads. Using tandem MS, they identified 204, 180, and 156 glycosylated proteins in day 14, 18, and 21 tumors, respectively; a majority of which were predicted to be membrane or secreted N-glycoproteins. Most recently, Chen and co-workers used this method to enrich glycoproteins in the mouse brain.²¹⁵

3.3.3. Labeling Fucosylated Glycoproteins with Fucose Analogues

Glycoprotein fucosylation is differentially regulated during normal and pathophysiological processes.²¹⁶ Fucose can be a terminal antenna glycan modification, as in the Lewis antigens where it is linked to an underlying GlcNAc or Gal residue, or a modification of the peptide-proximal GlcNAc residue within the N-glycan pentasaccharide core. As well, fucose directly linked to Ser or Thr side chains is the initiating residue of glycans associated with epidermal growth factor (EGF) and thrombospondin repeat (TSR) sequences, as in the EGF repeat-rich Notch protein.^217,218

To enrich fucosylated glycoproteins, both per-O-acetylated N-azidoacetylfucose (Ac₄FucAz) and per-O-acetylated N-alkynylacetylfucose (Ac₄FucAlk) have been used to metabolically label cellular glycans (see Figure 5b).^189,195,196 These compounds were designed to access the fucose salvage pathway, where free fucose is converted to fucose-1-phosphate and then to GDP-fucose, the substrate for all fucosyltransferases. However, fucose derivatives can have low labeling efficiencies and can inhibit protein fucosylation.²¹⁹ To overcome this limitation, Narimatsu and co-workers generated Namalwa cells that stably overexpressed two important enzymes in the fucose salvage pathway, L-fucokinase/GDP-fucose pyrophosphorylase (FKP) and α-1,3-fucosyltransferase 9 (FUT9).²²⁰ Overexpression of FKP and FUT9 enhanced metabolic labeling with Ac₄FucAlk and enabled the identification of FUT9 substrates.

Recently, FucAlk was used to label proteins bearing core O-fucose residues on their EGF or TSR domains, a modification catalyzed by protein-O-fucosyltransferases 1 and 2 (POFUT1 and POFUT2), respectively.²²¹ Through metabolic labeling, the O-fucose sites predicted from bioinformatics analysis were verified. Moreover, the unnatural fucose substrates were recognized by glycosyltransferases and elaborated analogously to the native fucose reside. This finding could enable the discovery of unexpected proteins bearing O-fucose residues.

3.3.4. Labeling Glycoproteins with GalNAc Analogues

GalNAc is a versatile monosaccharide that can be used to enrich both mucin-type O-glycoproteins and N-glycoproteins. In the case of mucins, a conserved α-GalNAc residue initiates all possible glycan core structures. GalNAc can also be found in different positions in N-glycan structures. Accordingly, GalNAc analogues that contain either an azide¹⁹² or alkyne¹⁹⁸ (Figure 5b) have been used to enrich a diversity of glycoproteins. Although some studies failed to distinguish the identifications into their respective glycoprotein class, others used lectins and/or PNGase F for N-glycoprotein assignment.

In an effort to identify prostate cancer biomarkers, Bertozzi and co-workers used metabolic labeling to enrich cell-surface glycoproteins from prostate cancer-derived PC3 cells.²²² Cells were fed Ac₄GalNAz and then treated with a membrane-impermeable phosphine-FLAG probe. By performing the bioorthogonal reaction on live cells, rather than cell lysates, the authors were able to select for cell-surface glycoproteins while minimizing potential contamination from intracellular glycoproteins (vide infra). After whole-cell lysates had been prepared, FLAG-labeled glycoproteins were enriched using an anti-FLAG antibody and then subjected to in-gel digestion with trypsin. Using tandem MS, the authors identified 71 proteins, 29 of which were previously known to be sialylated based on metabolic labeling with ManNAlk.²⁰³

In addition to the Staudinger ligation, Gorfien and coworkers used GalNAz and the CuAAC reaction to enrich glycoproteins from the CHO cell secretome.²²³ They cultured two different CHO cell lines, CHO-S and DG44, in Ac₄GalNAz and then treated the clarified cell culture media with an alkyne-functionalized agarose resin. Following on-bead digestion with trypsin, peptides were labeled with iTRAQ reagents, separated by two-dimensional chromatography, and quantified by tandem MS. The authors identified and quantified 325 secreted proteins between the two CHO cell lines.

To complement lectin- and periodate-based enrichment strategies, Smeekens et al. used metabolic labeling with GalNAc derivatives to enrich N-glycoproteins.²²⁴ They cultured HEK293T cells in Ac₄GalNAz and used copper-free click chemistry with a biotin-dibenzocyclooctyne probe to label cell-surface glycoproteins. After whole-cell lysates had been digested with trypsin, glycopeptides were enriched using NeutrAvidin beads and eluted by enzymatic deglycosylation using PNGase F. Following tandem MS, the authors identified 152 N-glycosites, which mapped to 110 glycoproteins. Although metabolic labeling with Ac₄GalNAz is agnostic to N- versus O-glycosylation, the authors restricted their analysis to N-glycoproteins by using PNGase F.

Recently, Xiao et al. evaluated metabolic labeling with three monosaccharide analogues for their ability to enrichment cell-surface N-glycoproteins.²²⁵ The authors demonstrated that GalNAz outperformed both ManNAz and N-azidoacetylglucosamine (GlcNAz) when combined with copper-free click chemistry for glycopeptide enrichment. The authors then used metabolic labeling with GalNAz and SILAC reagents to quantify changes in N-glycosites in statin-treated HepG2 liver cells.

3.3.5. Labeling O-GlcNAcylated Proteins with GalNAc Analogues

In addition to labeling cell-surface glycoproteins, GalNAz can also label intracellular O-GlcNAcylated substrates. Based on the promiscuity in sugar-donor selection by the enzyme UDP-galactose 4-epimerase (GALE), which interconverts UDP-GlcNAc and its C₄ epimer UDP-GalNAc, GalNAz can be converted to GlcNAz. Bertozzi and co-workers demonstrated this metabolic cross-talk by culturing Jurkat cells in Ac₄GalNAz and then treating whole-cell lysates with a phosphine-FLAG-His₆ probe.¹⁹³ After enrichment of glycoproteins by sequential pull-downs using an anti-FLAG antibody followed by metal affinity chromatography, the authors identified 18 proteins, many of which were previously known to be glycosylated. Depending on the application, the cross-talk between GalNAz and GlcNAz can require careful sample preparation to correctly interpret data from nucleocytosolic proteins and proteins translated in the ER/secretory pathway.^198,223

Recently, metabolic labeling with Ac₄GalNAz and the CuAAC reaction were combined for differential glycosylation analysis between two biological samples. The method is a modified version of difference gel electrophoresis (DIGE),^226,227 in which two samples are labeled with fluorophores that are distinct (but identical in charge and close in mass) and then combined and analyzed on a two-dimensional gel. After fluorescent gel spots that correspond to incongruity in the two fluorescent signals have been excised, protein abundance changes between the samples (i.e., due to a stimulus) can be identified through in-gel digestion and tandem MS. By integrating DIGE and metabolic labeling, Bertozzi and co-workers were able to rapidly monitor changes in glycoprotein levels in complex biological samples.²²⁸ Termed “glyco-DIGE”, this technique was used to examine organelle-specific protein O-GlcNAcylation. After Jurkat cells had been cultured in Ac₄GalNAz, mitochondrial and cytosolic extracts were separately labeled with an alkyne cyanine dye (either Cyanine 3 or Cyanine 5), combined, and then separated on a two-dimensional gel. Following in-gel digestion of fluorescent bands unique to the mitochondria, the authors used tandem MS to identify a mitochondrial outer-membrane protein with crucial roles in glucose metabolism and apoptosis regulation: voltage-dependent anion channel 2.

3.3.6. Labeling O-GlcNAcylated Proteins with GlcNAc Analogues

Intracellular O-GlcNAcylation is highly dynamic and typically substiochiometric, necessitating enrichment methods to identify O-GlcNAcylated proteins. To enrich this class of glycosylation, both per-O-acetylated N-azidoacetylglucosamine (Ac₄GlcNAz) and per-O-acetylated N-alkynylacetyl-glucosamine (Ac₄GlcNAlk) have been used for metabolic labeling (see Figure 5b).

In the first example, Zhao and co-workers fed Drosophila S2 cells Ac₄GlcNAz, treated cytosolic extracts with a phosphine-biotin probe, and captured labeled proteins on streptavidin beads.²²⁹ After proteins had been digested off the beads with trypsin and analyzed by tandem MS, the authors identified 41 proteins, 10 of which were previously known to be glycosylated. In a follow-up report, the same group applied an optimized enrichment protocol in HeLa cells and increased the number of identifications.²³⁰

In addition to the Staudinger ligation, the CuAAC reaction has also been used to enrich GlcNAz-labeled glycoproteins. For example, Kuster and co-workers treated whole-cell lysates from Ac₄GlcNAz-fed HEK293 cells with an alkyne-functionalized agarose resin.²³¹ Following on-bead digestion and tandem MS, they identified approximately 1500 proteins. For a subset of these identifications, the authors were able to resolve 185 O-GlcNAc glycosites using β-elimination (see section 4).

Lemonie and co-workers also used the CuAAC reaction to enrich glycoproteins. They treated nuclear and cytosolic extracts from Ac₄GlcNAz-fed MCF-7 cells with a biotin-alkyne probe.¹⁹⁹ After enrichment of glycoproteins with a streptavidin column and in-gel digestion with trypsin, the authors identified 32 proteins using tandem MS, 18 of which were previously known to be glycosylated. In the same work, Lemonie and co-workers tested metabolic labeling with Ac₄GlcNAlk, the alkyne-derivatized GlcNAc analogue. However, under their labeling conditions, they were unable to observe metabolic labeling.

In a later study, Pratt and co-workers also explored alkyne-functionalized GlcNAc derivatives for metabolic labeling and glycoproteomic profiling.¹⁹⁸ Under low-glucose growth conditions, they demonstrated that Ac₄GlcNAlk could be used for metabolic labeling in NIH3T3 cells. After whole-cell lysates were treated with a cleavable enrichment tag (azido-diazo-biotin), glycoproteins were captured onto streptavidin beads and released using sodium dithionite (which reductively cleaves the diazo bond). Using tandem MS, the authors identified 374 proteins, 95 of which were previously known to be glycosylated and three of which were confirmed by immunoprecipitation. The same authors also used this method to study glycoproteins in mice with diabetic retinopathy.²³²

One limitation of both Ac₄GlcNAz and Ac₄GlcNAlk is that enzymatic pathways can interconvert GlcNAc analogues into GalNAc analogues, necessitating cell fractionation and careful data analysis when assigning the glycosylation state of proteins post enrichment.^193,198 To address this limitation, Pratt and coworkers synthesized 6-azido-6-deoxy-N-acetylglucosamine (6AzGlcNAc), an unnatural monosaccharide that they found to be specific for O-GlcNAcylated proteins (Figure 5b).²³³ To validate the new GlcNAc analogue, the authors performed a comparative proteomics experiment by culturing NIH3T3 cells with peracetylated 6-azido-6-deoxy-N-acetylglucosamine (Ac₃6AzGlcNAc), Ac₄GlcNAz, or Ac₄GalNAz; treating each cell lysate with a biotin-alkyne probe; and enriching glycoproteins with streptavidin-conjugated beads. After eluting peptides by on-bead digestion and analyzing the mixture by tandem MS, they identified 366 proteins using Ac₃6AzGlcNAc, 350 of which were annotated as intracellular proteins.

3.3.7. Labeling Bacterial Glycoproteins with GlcNAc Analogues

Metabolic labeling with monosaccharide analogues was recently used to study glycoproteins in bacteria.²³⁴ Dube and co-workers used GlcNAz and the Staudinger ligation to enrich glycoproteins in Helicobacter pylori.^235,236 They cultured H. pylori in media supplemented with Ac₄GlcNAz for 3–5 days, treated whole-cell lysates with a phosphine-FLAG-His6 probe, and enriched glycoproteins by sequential pull-downs using an anti-FLAG antibody and metal affinity chromatography. Using multidimensional chromatography coupled with tandem MS, they identified 125 proteins, including a number of virulence factors. In the future, this strategy could be used to study bacterial glycosylation in different bacterial species.²³⁴

3.3.8. Advantages and Limitations of Metabolic Labeling

There are two key advantages to using metabolic labeling and the bioorthogonal chemical reporter strategy for glycoprotein enrichment. First, the chemical reporter is directly embedded into the glycan, allowing for glycosylation-specific enrichment using appropriate secondary probes. Unlike lectins or antibodies, which display weak binding and cross-reactivity, bioorthogonal reactions with azide- or alkyne-functionalized monosaccharides allow for stringent enrichment conditions. Second, unnatural monosaccharides can be dosed to cells or animals in a time-dependent manner (along with a stimulus) to reveal the dynamics and signal-dependent changes in glycosylation patterns. This is an appropriate strategy for identifying changes in low-abundance glycoproteins among a background of high-abundance steady-state protein glycosylation.

At the same time, there are limitations to the metabolic labeling strategy. First, some unnatural monosaccharides show low metabolic incorporation efficiency, as discussed with respect to the fucose analogues. This is likely the result of varying enzyme tolerances to substrate structural perturbations. Second, epimerase activity, which can convert one monosaccharide into another, should be characterized for each unnatural monosaccharide. Measurement of this activity would benefit from careful metabolic flux analysis, as was performed by Yarema and co-workers for ManNAc analogues.²³⁷ Third, unnatural monosaccharides could perturb cellular systems and might not be recognized with the same affinity by their wild-type binding partners. Finally, metabolic labeling strategies cannot be used to enrich glycoproteins from human samples.

3.4. Enrichment by Chemoenzymatic Tagging with Unnatural Monosaccharide Analogues

As an alternative to metabolic labeling, glycoproteins can be enriched after chemoenzymatic tagging with bioorthogonal chemical reporters.^174,176,177 In this approach, glycans are enzymatically appended with a functionalized monosaccharide and then covalently ligated to an enrichment probe using bioorthogonal reactions (Figure 6a). The challenge lies in finding or designing appropriate glycosyltransferases that can accept unnatural monosaccharide donors and act on their substrates extracellularly. The motivation for identifying such enzymes is cell-type-independent and high-efficiency labeling of glycoproteins.

Figure 6.

Enrichment of glycoproteins using chemoenzymatic tagging followed by bioorthogonal ligation. (a) Schematic showing how chemoenzymatic tagging can be used to introduce a chemical reporter directly onto glycoproteins in vitro. (1) UDP-galactose analogues are appended onto O-GlcNAc conjugates using a mutant galactosyl transferase (e.g., Y298L GalT). (2) A bioorthogonal chemical reaction is performed to label glycans with a probe molecule (red star). (b) Chemical structures of Y298L GalT substrates, including UDP-galactose (UDP-Gal) and two analogues, UDP-2-keto-Gal and UDP-GalNAz. (c) Structure of an azide-functionalized trisaccharide used to identify glycoprotein targets of a viral trans-sialidase.

3.4.1. Chemoenzymatic Tagging of O-GlcNAc

Chemo-enzymatic tagging of glycans was initially developed to improve the detection of O-GlcNAcylated proteins. It was first demonstrated by Hart and co-workers, who developed a radioassay using ³H-UDP-Gal and the enzyme β-1,4-galactosyl-transferase 1 (GalT1), which transfers Gal from UDP-Gal to any terminal GlcNAc residue.^102,238,239 Subsequently, Qasba and co-workers developed a nonradioactive detection method after they identified the key amino acids that governed GalT1’s sugar donor specificity.²⁴⁰ Using site-directed mutagenesis (i.e., Y289L), they enlarged the donor-substrate-binding pocket of GalT1, which enabled the enzyme to tolerate UDP-GalNAc and other GalNAc analogues at the 2-position. Hsieh-Wilson and co-workers leveraged the substrate tolerance of the Y289L GalT1 mutant enzyme, installing either GalNAz or the ketone isostere of GalNAc (2-keto-Gal) onto O-GlcNAcylated substrates (Figure 6b).^241–243

Hsieh-Wilson and co-workers conducted the first glycoproteomics study using the chemoenzymatic tagging enrichment strategy.²⁴⁴ They prepared nuclear and cytosolic extracts of rat brains, incubated the lysates with UDP-2-keto-Gal and the Y289L GalT1 mutant, and then treated the samples with an aminooxy biotin tag to label and enrich O-GlcNAcylated substrates. Using tandem MS, the authors identified 25 O-GlcNAcylated glycoproteins, two of which were previously known to be O-GlcNAcylated. In a follow-up study, Hsieh-Wilson and co-workers developed a quantitative variant of their chemoenzymatic tagging strategy.²⁴⁵ In this method, after chemoenzymatic tagging, glycopeptides were generated by trypsin digestion, and free amines were labeled using either isotopically light or heavy versions of formaldehyde and sodium cyanoborohydride. The authors used this protocol to quantify the dynamics of O-GlcNAcylation in the rat brain under different stimulatory conditions. Later, Hart and co-workers used the same enrichment strategy to identify 35 O-GlcNAc sites on 25 glycoproteins in human erythrocytes.²⁴⁶

In addition to oxime-forming reactions, the CuAAC reaction has also been used with chemoenzymatic tagging. In the first example, Hsieh-Wilson and co-workers incubated rat forebrain lysates with UDP-GalNAz and the Y289L GalT1 mutant enzyme to install azides onto O-GlcNAcylated substrates.²⁴¹ After treatment with a tetramethyl-6-carboxyrhodamine-(TAMRA-) alkyne probe and enrichment by immunoprecipitation, the authors identified 213 proteins using tandem MS, 67 of which were previously known O-GlcNAcylated proteins. Later, Palmisano and co-workers used the same strategy to install azides onto O-GlcNAcylated substrates in mouse brain lysate. The authors then treated the labeled lysate with a phospho-alkyne tag and used titanium dioxide chromatography to simultaneously enrich for phosphopeptides and glycopeptides.²⁴⁷ Using tandem MS, they identified 12 O-GlcNAcylated peptides and 485 phosphopeptides. Subsequently, after optimizing conditions to favor glycopeptide enrichment, the authors identified 20 O-GlcNAcylated peptides and 23 phosphopeptides. Additionally, Hart and co-workers used chemoenzymatic tagging to determine that histones were O-GlcNAcylated.²⁴⁸

To improve the recovery of enriched O-GlcNAcylated substrates, chemoenzymatic tagging was followed by treatment with cleavable enrichment probes. In the first example, Hart and co-workers synthesized a photocleavable biotin-alkyne probe to react with GalNAz-tagged O-GlcNAcylated peptides from rat brain lysates. After affinity enrichment using avidin chromatography, the authors released glycopeptides through exposure to UV light (365 nm) and identified them using tandem MS.²⁴⁹ The photocleavage reaction reduced the mass of the tagged peptide and generated a basic aminomethyl-triazole group, both favorable features for tandem MS. Hart and co-workers employed this strategy in two large-scale O-GlcNAc profiling studies: One investigated the interplay between O-GlcNAcylation and phosphorylation during mitosis (141 glycosites in 64 glycoproteins),²⁵⁰ and the other investigated differences in O-GlcNAcylation in cerebrocortical brain tissue between wild-type mice and an Alzheimer’s disease model (458 glycosites in 274 glycoproteins).²⁵¹

More recently, three other cleavable enrichment probes have been combined with chemoenzymatic tagging of O-GlcNAcylated substrates. In one example, Tanaka and co-workers used a 3-ethynylbenzaldehyde probe for reversible hydrazone formation with a hydrazide resin so that enriched glycopeptides could be eluted using hydroxylamine.²⁵² In another example, Tsumoto et al. used a thiol-alkyne probe for reversible disulfide formation with a thiol-reactive resin.²⁵³ Finally, Hsieh-Wilson and co-workers used an alkyne-functionalized 1-(4,4-dimethyl-2,6-dioxocyclohex-1-ylidene)ethyl biotinylated probe that can be quantitatively cleaved by hydrazine.²⁵⁴

3.4.2. Chemoenzymatic Tagging of Sialic Acid

In host–pathogen interactions, sialic acid residues from the pathogen can be transferred to the host’s cell surface through a trans-sialidase enzyme.²⁵⁵ In studies of this behavior, it has been difficult to identify cell-surface glycoproteins on host immune cells that act as acceptors for sialic acid residues originating from the pathogen, so Campetella and co-workers developed a chemozymatic tagging strategy to study this interaction.^255,256 They characterized the substrate preference of a trans-sialidase from Trypanosoma cruzi using an azide-functionalized sialic acid derivative, SiaNAz-α2,3-Lac-β-OMe (Figure 6c).²⁵⁵ To demonstrate the strategy, they incubated Jurkat cells with the microbial trans-sialidase enzyme and the SiaNAz derivative and then performed the Staudinger ligation using a phosphine-FLAG probe on live cells. Cell lysates were prepared, and FLAG-labeled glycoproteins were enriched by immunoprecipitation with an anti-FLAG antibody. Using tandem MS, the authors identified CD45 isoforms as the main glycoprotein acceptor of the SiaNAz derivative.

3.4.3. Chemoenzymatic Tagging of Disaccharides

To extend the technique of chemoenzymatic tagging to other glycans, new enzymes with appropriate target specificity and substrate tolerances are required. This effort has been used to target two disease-relevant disaccharides: N-acetyllactosamine (LacNAc, Gal-β-1,4-GlcNAc) and fucose-α-1,2-galactose (Fuc-α-1,2-Gal). In the case of LacNAc, Qasba and co-workers designed a mutant bovine α-1,3-galactosyltransferase that could transfer GalNAz or 2-keto-Gal onto LacNAc residues.²⁵⁷ In another approach by Wu and co-workers, the authors demonstrated that wild-type α-1,3-fucosyltransferase from H. pylori could transfer fucose analogues (e.g., 6-azido-fucose) onto LacNAc residues.²⁵⁸ Finally, to target Fuc-α-1,2-Gal, Hsieh-Wilson and co-workers demonstrated that the bacterial glycosyltransferase BgtA could transfer GalNAc analogues (e.g., 2-keto-Gal or GalNAz) onto the Gal residue within the Fuc-α-1,2-Gal motif. In the future, we anticipate that these new chemoenzymatic tagging strategies will be coupled with enrichment tags to help identify glycoprotein substrates containing higher-order glycans.

3.4.4. Advantages and Limitations of Chemoenzymatic Tagging

The main advantage to using the chemoenzymatic tagging strategy for labeling glycoproteins with bioorthogonal chemical reporters is the high efficiency of the reaction. At the same time, there are limitations to the method. First, chemoenzymatic tagging is best suited to report on steady-state glycosylation rather than on its dynamics. Second, designing enzymes to tolerate unnatural monosaccharide substrates and synthesizing and purifying sugar derivatives (e.g., UDP-sugars) can be difficult.²⁵⁹

4. IDENTIFICATION OF SITES OF GLYCOSYLATION

4.1. Overview

The complete characterization of a proteome includes identifying the locations and the structures of PTMs on proteins. Tandem MS has become an invaluable tool for this task, although considerable challenges exist for the unambiguous assignment of glycosylation sites and glycan structure.²⁶⁰ This is particularly true in cases involving low protein abundance or low, substoichiometric site occupancy. Glycopeptides also have diverse ionization efficiencies, dependent on both the type of glycan and the type of mass spectrometer, affecting data quality and detectability by MS.^261,262 Adding to the experimental challenges, software tools for analyzing mass spectra of intact glycopeptides are limited, with few solutions for accurately identifying the site of modification and the glycan composition, and for providing confidence metrics for the assignments.²⁶³

In this section, we discuss how chemical techniques and MS can help identify sites of glycosylation in complex protein mixtures; for more information on characterizing glycan structures, see recent glycomics reviews.^{28,69,88–91} In one approach, enzymatic or chemical deglycosylation methods are used to “mark” the glycosite, enabling indirect identification of the glycosite using traditional tandem MS techniques with the deglycosylated peptide. In another approach, advances in tandem MS fragmentation methods [e.g., electron-transfer dissociation (ETD)] are utilized for the direct analysis of intact glycopeptides. In the final approach, homogeneous and well-defined glycans are generated to reduce the computational challenges that typically accompany the analysis of large and heterogeneous glycopeptides.

In parallel with glycosite mapping and glycomics, there are techniques that characterize intact glycopeptides to provide both the site of modification and the glycan composition. These techniques leverage improvements in mass spectrometers and computational tools to simultaneously provide peptide and glycan sequence information. Although we discuss some studies using these techniques in the context of chemical and enzymatic methods, for more information on large-scale intact glycopeptide studies, see recent reviews by Thaysen-Andersen et al.^86,87 and Nilsson,¹⁰⁹ as well as recent work by Medzihradszky and co-workers.^264,265

4.2. Enzymatic and Chemical Deglycosylation

Deglycosylation methods can improve glycosite mapping in a number of ways. First, after the glycan has been removed, peptides can be identified using standard tandem MS methods with collision-induced dissociation (CID). In the absence of deglycosylation, CID of an intact glycopeptide typically results in glycan cleavage and fragmentation, leaving the peptide intact. Although it is possible to optimize CID settings to improve the peptide backbone fragmentation of intact glycopeptides, a mixture of carbohydrate side-chain and peptide cleavage events typically occur.^78,266–269 With instruments that allow for nested fragmentation (i.e., MSⁿ), CID fragmentation of the peptide ion resulting after glycan loss (i.e, MS³) can be performed to obtain peptide sequence information.^109,165,270 However, the decrease in fragmention abundance with sequential stages of tandem MS can limit this approach. Second, deglycosylation can improve the proteolytic digestion efficiency by increasing enzymatic access to the amino acid backbone. For these reasons, deglycosylation has become an important step in the identification of both N- and O-glycosites.

4.2.1. Deglycosylation Methods for N-Glycosylation

The most common method for mapping N-glycosites relies on proteomic analysis of enzymatically deglycosylated N-glycopeptides. PNGase F is an amidase that cleaves the entire glycan off an asparagine residue and simultaneously converts the residue into aspartic acid (Asp).²⁷¹ This enzymatic deamidation process marks the previously glycosylated Asn residue with a 0.98-Da mass increase, which can be detected using high-resolution tandem MS measurements (Figure 7a).²⁷² To improve the deglycosylation efficiency of N-terminal Asn residues, Weng et al. succinylated the N-terminus before PNGase F treatment.²⁷³

Figure 7.

Examples of enzymatic and chemical deglycosylation. (a) N-glycans can be enzymatically removed using N-glycosidases or endoglycosidases, most commonly PNGase F (top), which converts asparagine into aspartic acid after cleaving off the entire glycan. If performed in the presence of ¹⁸O-labeled water, the heavy oxygen (highlighted in red) will mark the site of glycosylation. Alternatively, Endo H (bottom) cleaves between the two innermost GlcNAc resides, leaving a single peptide-bound GlcNAc residue to mark the site of glycosylation. (b) O-glycans can be chemically deglycosylated. Illustrated is a schematic of the β-elimination followed by Michael addition with dithiothreitol (DTT) (BEMAD) method. (1) β-elimination, most commonly with NaOH, converts glycosylated serine or threonine residues into dehydroalanine or dehydrobutyric acid, respectively. (2) Michael addition with amine or thiol nucleophiles, most commonly DTT, covalently marks the site of glycosylation.

To further increase the confidence in N-glycosite assignment, Kaji et al.²⁷⁴ developed a strategy based on stable isotope labeling.²⁷⁵ The authors performed PNGase F deglycosylation in the presence of isotopically enriched ¹⁸O-water, imparting a 2.98-Da mass increase following the Asn-to-Asp conversion. Termed isotope-coded glycosylation site-specific tagging, the method was used to identify N-glycosites in Caenorhabditis elegans and was the first strategy to produce a large-scale map of N-glycosites.²⁷⁴ Similar strategies have been used to identify N-glycosites in other organisms, including mice⁹⁸ and Trypanosoma cruzi.²⁷⁶ One limitation to this method is that the spontaneous deamination of nonglycosylated Asn residues during the course of PNGase F treatment can yield false positive identifications.^277–279

To reduce the ambiguities in N-glycosite mapping that can accompany complete glycan removal, an alternative approach relies on trimming N-glycans to a single carbohydrate residue. For example, the enzyme endoglycosidase H (Endo H) preserves a single peptide-proximal N-linked GlcNAc residue after treatment (Figure 7a).^280,281 These truncated N-glycopeptides, like their completely deglycosylated counterparts, can be sequenced by tandem MS using standard CID conditions because the N-glycosidic bond is stable to CID fragmentation.²⁸⁰ Unfortunately, the presence of a single monosaccharide can suppress ionization, potentially biasing data-dependent tandem MS data acquisition against glycopeptide ions.^261,262 Advances in targeted proteomics can be used to overcome this limitation (see section 5).

4.2.2. Deglycosylation Methods for O-Glycosylation

Unlike for N-glycans, there is an absence of universal enzymes that can globally deglycosylate all O-glycan structures. To circumvent this limitation, chemical deglycosylation strategies have been developed. One method relies on periodate oxidation of glycans followed by treatment with a base to sequentially eliminate carbohydrates until just the amino acid backbone remains.^282–284 Another approach, hydrazinolysis, utilizes anhydrous hydrazine to release either all glycans or O-glycans selectively, depending on the specific reaction conditions.²⁸⁵ Another strategy uses anhydrous trifluoromethanesulfonic acid (TFMSA) for the partial or complete deglycosylation of either N- or O-glycans, respectively.²⁸⁶ When TFMSA is used at 0 °C, peripheral carbohydrates are eliminated, whereas the peptide-proximal monosaccharide is preserved (i.e., preservation of O-GalNAc for mucins or N-GlcNAc for N-glycans). However, when it is used at higher temperatures (25 °C or greater), complete deglycosylation of entire O-glycans is possible.²⁸⁷ Typically, when TFMSA is used, scavengers such as anisole are added to the reaction mixture to minimize degradation of the peptide backbone.²⁸⁷ Collectively, these chemical deglycosylation methods have primarily been used to identify glycosites on single, purified glycoproteins or a defined glycopeptide mixture; their compatibility with large-scale glycosite mapping studies remains to be demonstrated.

In contrast, β-elimination of O-glycans is one chemical deglycosylation method that has found its way into large-scale glycosite mapping studies. It relies on the base-mediated conversion of glycosylated Ser or Thr residues into dehydroalanine or dehydrobutyric acid, respectively (Figure 7b). The resulting change in the mass of the amino acid can be used to assign O-glycosites in tandem mass spectra.^288,289 Initially, this method was used to identify O-glycosites in simple mixtures of glycopeptides or in purified mucin-type glycoproteins such as MUC1.^290–292 The strategy was later modified by adding amine or thiol nucleophiles to react with the generated α,β-unsaturated carbonyl through a Michael addition. By simultaneously eliminating the glycan and replacing it with a chemical tag that could enable enrichment and was stable to CID fragmentation, this method was quickly applied to a number of glycosite mapping studies. At first, reagents that could act as both a base and a nucleophile were used, including methylamine, ethylamine, and dimethylamine. Later, a two-step variant was developed using sodium hydroxide as the base followed by dithiothreitol (DTT) as the nucleophile. Termed β-elimination followed by Michael addition with dithiothreitol (BEMAD),²⁹³ the method was expanded to include a variety of nucleophiles, including butanethiol, aminoethanethiol, and biotin-pentylamine or biotin-cystamine for enrichment of the deglycosylated peptides (Figure 7b).^{244,246,294,295} The BEMAD method has primarily been used to identify sites of O-GlcNAcylation in samples ranging in complexity from purified protein complexes and cellular organelles to mammalian whole-cell lysates.^{231,248,262,296} By using isotopically light DTT (DTT-d₀) or heavy DTT (DTT-d₆) during the Michael addition step, differential isotopic labeling for comparative and quantitative analysis of glycosites is possible.²⁹⁷ Hart and co-workers used this strategy to characterize changes in protein O-GlcNAcylation in cardiac mitochondria isolated from rats treated with or without Thiamet G, an inhibitor of O-GlcNAcylation.²⁹⁸ Using a similar approach, Hart and co-workers explored the role of O-GlcNAcylation in the mitochondria of diabetic rat hearts.²⁹⁹

Despite the potential utility of chemical deglycosylation in glycosite mapping, this method suffers from a number of drawbacks. Most notably, the harsh conditions that are sometimes needed to promote deglycosylation (e.g., high concentrations of acids and bases at elevated temperatures) can affect the peptide backbone, leading to nonspecific cleavage and amino acid modifications. Additionally, because other O-linked PTMs are labile under acidic and basic conditions (e.g., phosphorylation under basic conditions), appropriate controls and careful sample preparation are necessary to correctly interpret the MS data.

4.3. Electron-Transfer Dissociation

Advances in MS fragmentation technology have improved the characterization of labile PTMs. These methods can be used to directly characterize intact glycopeptides, increasing the confidence of glycosite assignments. In particular, electron-capture dissociation (ECD)^300,301 and electron-transfer dissociation (ETD),^77,302,303 which do not use high collisional energies to fragment ions (as does CID), often preserve labile PTMs. Consequently, ECD and ETD preferentially fragment the peptide backbone of glycopeptides, enabling the identification of both the peptide and the site of modification.^78,304–307 Using both ECD and ETD, N- and O-glycosites have been identified in a number of biological samples.⁶⁴

Although both ETD and ECD have been used for glycosite identification, ETD gained popularity after it was integrated with ion-trap mass spectrometers (e.g., LTQ-Orbitrap).³⁰⁸ These instruments are low-cost alternatives to the Fourier transform ion cyclotron resonance mass spectrometers that are necessary for ECD fragmentation. In particular, ETD has emerged as the fragmentation method of choice for globally mapping sites of O-GlcNAcylation. When implemented, this method has required glycopeptide enrichment through the use of antibodies,³⁰⁹ LWAC,^{103,310–313} or chemoenzymatic methods,^{245,249,250,262} followed by ETD-based tandem MS. ETD has also been used for O-GlcNAc glycosite mapping in Campylobacter jejuni whole-cell lysates,³¹⁴ in the murine synaptosome,³¹⁵ in mouse liver and brain tissue,²⁶⁵ and in Mycobacterium tuberculosis culture filtrates.³¹⁶ Most recently, ETD was used to quantify the dynamics in O-GlcNAcylation sites after metabolically labeling HeLa cells with ¹³C₆-glucose (which was converted to ¹³C-labeled UDP-GlcNAc).³¹⁷

Compared to CID, precursor-ion fragmentation by ETD (and ECD) is less efficient and best suited for high charge state ions (Z > 3) with m/z values less than 1000.^318,319 Accordingly, samples that contain glycopeptides with trimmed glycans and that ionize as high charge state species are ideal for ETD analysis.^64,68 Analysis and interpretation of ETD mass spectra for glycopeptides is also challenging. Darula et al. demonstrated that the number of identified glycopeptides and glycosites could be more than doubled by reprocessing an ETD data set using a new version of Protein Prospector that implemented a modified scoring algorithm.^320,321 The authors also identified side reactions between sialic acid and common enrichment buffers, leading to additional complications during data analysis.³²² In some cases, an iterative data analysis strategy is necessary, where a range of glycan masses are considered as variable modifications on Asn, Ser, or Thr residues. Then, the most likely glycan structure(s) can be estimated based on the measured mass of the modification.³¹⁵

4.4. Well-Defined and Homogeneous Glycans

Mucin-type O-glycosylation has been notoriously difficult to study by MS because of the heterogeneity in O-glycan structures and the high density of adjacent glycosylation sites. To reduce sample complexity, early attempts at O-glycosite mapping required a single, purified glycoprotein; this strategy was employed for MUC1, β-amyloid precursor protein, immunoglobin A1, and lubricin.^{54,68,323–326} For complex mixtures, where the interpretation of mucin-type glycopeptide tandem mass spectra is challenged by a large diversity of glycan structures, two general strategies have been used to address the issue: (1) lectins to enrich for mucin populations that express the same or similar glycans and (2) genetically engineered cell lines to produce homogeneous glycans on all mucins.

4.4.1. Well-Defined Glycans through Lectin Enrichment

Lectins have been used to enrich for proteins or peptides containing specific O-glycan structures. In a workflow developed by Medzihradszky and co-workers, the lectin Jacalin was used to enrich for O-glycopeptides from a bovine serum digest, and 26 O-glycosites were identified using ETD.^321,327 Subsequently, the same group improved the performance of this workflow by coupling additional chromatography steps after lectin enrichment. Using the new method, the authors identified 124 O-glycosites across 51 glycoproteins.³²⁸ A similar lectin enrichment strategy was also employed by Bagdonaite et al. to identify O-glycosites on viral envelope proteins in herpes simplex virus type 1.³²⁹

4.4.2. Homogeneous Glycans through Genetic Engineering

Clausen and co-workers developed a genetic engineering strategy to identify and study O-glycosites in mucins (see recent review³³⁰). They generated a cell line, termed “SimpleCells”, in which they silenced the chaperone protein cosmic and produced truncated but homogeneous O-glycans (either Tn or STn) on mucin-type glycoproteins.^331,332 Using lectin affinity chromatography and ETD fragmentation, the authors were able to enrich the truncated mucins and identify O-glycosites. Recently, they improved the enrichment protocol through prefractionation of samples with isoelectric focusing, proteolysis with chymotrypsin, and a two-stage LWAC enrichment step.³³³

Using the SimpleCell technology, Clausen and co-workers identified several thousand O-glycosites across a number of cell lines, including CHO cells and 15 human cancer cell lines (HEK293, HeLa, OVCAR-3, MCF-7, MDA-MB-231, T47D, IMR-32, K562, HepG2, HaCaT, COLO-205, Capan-1, AGS, MKN45, and T3M4).^{331,334–338} Clausen and co-workers also used the SimpleCell technology to identify substrates for specific glycosyltransferases, focusing on ppGalNAcT isoforms. They performed a comparative glycoproteomics experiment between wild-type and ppGalNAcT2 knockout SimpleCells and identified 73 unique sites of glycosylation for the ppGalNAcT2 isoform.³³⁹

Beyond mucin-type O-glycosylation, SimpleCells have also been used to identify O-mannosylated glycoproteins. Through the silencing of both cosmic and the glycosyltransferase POMGnT1 in MDA-MB-231 cells, mucin glycans were truncated to a single GalNAc residue, and mannosylated glycans were truncated to a single mannose residue.³⁴⁰ Using this strategy, Clausen and co-workers identified 235 O-Man glycosites, mapping to 51 glycoproteins. Recently, when this method was extended to study O-mannosylated glycoproteins in yeast, the authors discovered O-Man glycosites on nuclear, cytoplasmic, mitochondrial, and secreted proteins.^341,342

Although SimpleCells eliminate knowledge about site-specific glycan composition (i.e., microheterogeneity), they provide a path toward large-scale O-glycoproteomics and O-glycosite mapping. The volume of data generated by these studies has already improved computational models for predicting sites of O-glycosylation in a given protein sequence (e.g., NetOGlyc).³³⁶

5. TARGETED GLYCOPROTEOMICS

5.1. Overview

In targeted proteomics, the mass spectrometer is instructed to detect specific ions of interest, enabling high sensitivity, reproducibility, and quantitative accuracy.^343,344 Popular implementations include selected reaction monitoring (SRM)/multiple reaction monitoring (MRM) and parallel reaction monitoring (PRM).^345,346 This technology has enabled hypothesis-driven research in systems biology and clinical studies for biomarker discovery, among other applications.³⁴⁷ Critical to the success of targeted proteomics is the development of repositories containing lists of target proteins, identification of proteotypic peptides, and characterization of a peptide’s physical properties to facilitate MS-based measurements.³⁴⁵ This effort has generated a number of public resources for targeted proteomics, including Skyline, the SRMAtlas, and the CPTAC Assay portal.^348–350 Recently, targeted proteomics was extended to glycoproteins through MRM-based quantitation of glycoforms,³⁵¹ PRM assays for N-glycosites,³⁵² and the introduction of an N-glycosite SRMAt-las.³⁵³ The SRMAtlas was critical in the analysis of human breast tumor samples to identify potential N-glycoprotein biomarkers.³⁵⁴ A related approach, relying on data-independent acquisition and targeted-data analysis [e.g., sequential window acquisition of all theoretical fragment ion spectra MS (SWATH-MS)], has also been applied to the identification of N-glycosites.¹⁴⁸ Using human tumor tissue samples, Liu et al. employed SWATH-MS to identify N-glycoprotein biomarkers that correlated with the level of aggressiveness of prostate cancer.³⁵⁵ SWATH-MS has also been used to study N-glycosite micro- and macroheterogeneity in yeast cell wall glycoproteins.^356,357

Inspired by the benefits of targeted proteomics, Bertozzi and co-workers developed a related technology based on chemically tagging proteins of interest.³⁵⁸ Once tagged, peptides are detected and sequenced in a manner similar to, but distinct from, that employed in targeted proteomics. This strategy, generally termed chemically directed proteomics, has been used to identify both N- and O-glycosites and glycoproteins in both yeast and human cells.^359,360

5.2. Chemically Directed Glycoproteomics

A common challenge in proteomics experiments is identifying low-abundance proteins in complex biological samples. This is especially true in studies of glycoproteins, where glycans can suppress ionization. To address this problem, Bertozzi and coworkers developed a chemically directed proteomics strategy for identifying low-abundance species in complex mixtures. The method was quickly extended to study glycoproteins and proved valuable in both resolving the site of glycosylation and sequencing complex, intact glycopeptides. Termed isotopic signature transfer and mass pattern prediction (IsoStamp), the technique exploits the perturbing effects of a dibromide motif on a peptide’s isotopic envelope³⁵⁸ and is an extension of isotopic distribution encoding tagging.³⁶¹ As illustrated in Figure 8a, the isotopic envelope of an average peptide changes significantly upon modification with two bromine atoms. Two bromine atoms produce a symmetrical triplet in a mass spectrum, with major peaks at M, M + 2, and M + 4 at a relative intensity ratio of 1:2:1 due to the abundances of the ⁷⁹Br₂, ⁷⁹Br⁸¹Br, and ⁸¹Br₂ isotopic pairings. The isotopic envelope of the dibromide-tagged peptide then reflects the parent peptide’s intrinsic isotopic distribution convoluted with (or “recoded” by) the dibromide distribution.

Figure 8.

Chemically directed glycoproteomics using chemical or metabolic methods to embed a dibromide or dibromide-like motif into glycans. (a) Two bromine atoms perturb a peptide’s isotopic envelope in a predictable way. Illustrated are simulated isotopic envelopes of an AVERAGEPEPTIDE, a dibromide (Br₂), and a dibromide-labeled AVERAGEPEPTIDE, the last of which can be computationally recognized with high sensitivity and specificity in complex mass spectra. (b) Chemical structure and mass spectrum of an iodoacetamide-derivatized dibromide tag that was used to chemically modify peptides by alkylating cysteine residues. (c) Chemical structure of a cleavable dibromide enrichment tag for isotope-targeted glycoproteomics. The silane linkage can be cleaved using 2% formic acid, producing a low-molecular-weight tag that remains on the labeled glycopeptide. (d) A GlcNAc isomix was used to metabolically install a dibromide-like motif directly into glycoproteins for high-confidence glycosite mapping without chemical tagging. The GlcNAc isomix consisted of three components and was designed to mimic the 1:2:1 peak intensity distribution of the dibromide triplet pattern using heavy nitrogen and carbon instead of bromine atoms.

In the first example of this technology, the dibromide motif was installed onto peptides using a dibrominated tyrosyl iodoacetamide analogue capable of alkylating cysteine residues (Figure 8b).³⁵⁸ This labeling strategy was used in a proof-of-principle study to develop a computational pattern-searching algorithm capable of identifying isotopically recoded peptides in complex samples with high sensitivity and fidelity. Following pattern identification, the authors used a directed proteomics strategy to identify dibromide-labeled peptides, where inclusion lists containing m/z values and retention times of precursor ions bearing a recoded isotopic envelope were used to trigger tandem MS. Thus, the IsoStamp approach utilizes isotopic pattern rather than ion abundance to drive ion selection for fragmentation. This technique outperforms a traditional data-dependent acquisition strategy, particularly at low target-analyte concentrations.

Subsequently, Bertozzi and co-workers applied the IsoStamp technology to glycoproteins by introducing isotope-targeted glycoproteomics (IsoTaG).^359,362 This method involves metabolic labeling of cells with unnatural monosaccharide analogues, chemical tagging using a dibrominated affinity probe, and chemically directed proteomics to enable fragmentation and identification of glycopeptides. Importantly, IsoTaG is compatible with whole-proteome-scale profiling of intact N-and O-glycopeptides, providing information on both the peptide sequence and glycan structure. The authors used three cancer cell lines (Jurkat, PC-3, and MCF-7) and two unnatural monosaccharide analogues (Ac₄GalNAz and Ac₄ManNAz) to generate a diversity of cell lysate samples. The lysates were treated with a cleavable, dibrominated, biotin-alkyne probe (Figure 8c) that was used to enrich for glycoproteins using streptavidin-agarose beads. After peptides had been eluted by on-bead digestion, glycopeptides were separately enriched by treatment with formic acid to cleave the acid-labile silane linkage in the enrichment probe. Subsequently, peptides were identified using a traditional data-dependent acquisition strategy, and glycopeptides were analyzed using a modified version of the directed proteomics strategy described above. By leveraging the capabilities of an ion-trap mass spectrometer, the authors combined MS²- and MS³-level CID information to obtain both glycan and amino acid sequence information. Collectively, the authors identified 32 intact N-glycopeptides and over 500 intact O-glycopeptides from 250 glycoproteins, including multiple glycoforms for some glycopeptides. Recently, IsoTaG was also used to study the role of the glycosyltransferase GALNT3 in an epithelial ovarian cancer cell line.³⁶³

Although the isotopic signature of a halogenated chemical tag can effectively highlight labeled peptides or glycopeptides within a complex LC-MS data set, some biomolecules are not amenable to chemospecific modification. To address this limitation and expand the utility of the IsoStamp platform, Bertozzi and co-workers demonstrated that it was possible to recode a peptide’s isotopic envelope without chemical tagging. Instead, a dibromide-like isotopic signature was embedded into glycans through metabolic labeling.³⁶⁰ To achieve this effect, a stoichiometrically defined mixture of GlcNAc isotopologues, referred to as a GlcNAc isomix (Figure 8d), was prepared by combining three GlcNAc isotopologues in a 1:2:1 molar ratio (i.e., to mimic the 1:2:1 dibromide intensity ratio). To evaluate the ability of the GlcNAc isomix to recode the peptide’s isotopic envelope, the isomix was fed to a mutant yeast strain with an engineered salvage pathway for GlcNAc.³⁶⁴ The isomix was metabolically installed into all yeast N-glycans by virtue of the fact that eukaryotic N-glycans contain a conserved GlcNAc-β1,4-GlcNAc disaccharide at their reducing end (see Figure 1a). After metabolic labeling, yeast lysates were prepared and digested with either trypsin or chymotrypsin, and glycopeptides were enriched using a modified version of FASP.^69,168 During the process, N-glycans were truncated to a single GlcNAc residue with Endo H, which has a preference for high-mannose-type oligosaccharides. This strategy preserved the isotopic signature of the GlcNAc isomix and the site of glycosylation. Glycopeptides bearing the isomix signature were subjected to a directed proteomics workflow in which precursor ions were selected for CID using a 4-Da isolation window (IW). A broad IW was used to preserve the isomix signature in fragment ions bearing the GlcNAc isomix modification, further increasing confidence in glycosite assignments. Using this strategy, the authors identified 133 N-glycosites spanning 58 glycoproteins within the yeast proteome, 50% of which were previously biochemically validated as N-glycoproteins.

Chemically directed glycoproteomics is a versatile platform for characterizing glycoproteins in complex biological samples. The IsoStamp technology enables the detection of low-abundance species and offers an enhanced level of confidence for mapping glycosites. With the introduction of IsoTaG, whole-proteome-scale profiling of intact N- and O-glycopeptides is feasible. At the same time, there are limitations to the technology. In particular, the isotopic tagging method dilutes the signal intensity of the parent ion across multiple isotope peaks.³⁵⁸ Additionally, halogenated glycopeptides can have low ETD fragmentation efficiency, necessitating nested CID fragmentation (MS²/MS³) for glycopeptide sequencing.^359,362 However, with continued development, this strategy could be adapted to other chemistries and could be used to profile PTMs beyond glycosylation.

6. FUNCTIONAL GLYCOPROTEOMICS

6.1. Overview

An important aspect of connecting glycoproteins to cellular functions is building interaction networks that can delineate the connectivity between a glycoprotein and its binding partners. One approach to constructing these maps is to stabilize glycoprotein interactions with chemical cross-linking probes and then identify the interacting proteins with MS. Unfortunately, most chemical cross-linking probes target amino acids and ignore glycosylation-dependent interactions.³⁶⁵ These interactions should also be studied in a physiologically relevant environment to capture biologically relevant proteins.³⁶⁶ In this section, we discuss methods for studying glycoprotein interactions in a glycosylation-dependent manner and in the context of living cells.

6.2. Capturing Cell-Surface Glycoprotein Receptors

Wollscheid and co-workers developed a method for identifying the cell-surface glycoprotein receptor for a given protein ligand.³⁶⁷ Termed ligand-based receptor-capture (LRC) technology, this method builds on cell-surface capturing technology (see section 3.2.2).¹⁵⁵ The LRC technology uses a trifunctional chemoproteomics probe called TRICEPS (Figure 9a) that contains an N-hydroxysuccinimide ester for conjugation to a ligand protein, a trifluoroacetylated hydrazine for condensation with aldehydes introduced into glycans through oxidation, and a biotin moiety for affinity enrichment. After the TRICEPS probe was conjugated to a ligand of interest, it was incubated with sodium periodate-treated cells or tissue samples. Given that many cell-surface proteins are glycosylated, ligand–receptor interactions were stabilized through covalent capture between the hydrazine functionality of the TRICEPS probe and aldehydes introduced on oxidized glycans in cell-surface glycoproteins. After labeling, the cells were lysed, digested with trypsin, and the TRICEPS-modified glycopeptides were enriched with streptavidin beads. Following the release of N-glycopeptides by PNGase F treatment and analysis by quantitative MS, the enrichment of a receptor, relative to a negative control sample, indicated a true ligand–receptor interaction. A number of well-characterized ligands, including insulin, transferrin, apelin-17, epidermal growth factor, and the antibody trastuzumab, were conjugated to the TRICEPS probe, and in each case only the known cell-surface receptor was identified. The TRICEPS probe was used to identify the cell-surface receptors targeted by vaccinia virus. Collectively, the LRC technology provides a route to identify specific ligand–glycoprotein receptor interactions.

Figure 9.

Chemical tools for covalently capturing glycoproteins and glycan-binding proteins. (a) Chemical structure of the trifunctional chemoproteomics (TRICEPS) probe that was used to capture cell-surface glycoprotein receptors for a specific ligand. (b) Chemical structures of sialic acid derivatives bearing an aryl azide photo-activatable cross-linking functional group. (c) Chemical structures of sialic acid derivatives bearing a diazirine photoactivatable cross-linking functional group. (d) Chemical structures of GlcNAc derivatives bearing a diazirine photoactivatable cross-linking functional group.

6.3. Identifying Glycan-Binding Proteins

Glycan-binding proteins (GBPs) are difficult to isolate using traditional enrichment techniques, because they typically have low binding affinities and the complexes they form can be short-lived.³⁶⁸ One strategy to access these complexes relies on the formation of covalent interactions between the glycan and the binding protein using photo-cross-linking reagents.³⁶⁹ Early work in this field focused on the synthesis of glycans, glycopeptides, and glycolipids containing both a photo-cross-linking group and an enrichment moiety (see recent reviews by Kohler and co-workers for more information).^24,376 These reagents were used to enrich and identify a variety of GBPs from cell culture media, including the asialoglycoprotein receptor, the adhesion factor BabA on H. pylori, and the B-subunit of cholera toxin.^370–372

Another way to use photo-cross-linking reagents to identify GBPs is through metabolic incorporation of photo-cross-linker-functionalized monosaccharides (Figure 9b,c).^366,369 For example, two sialic acid analogues bearing an aryl azide photo-cross-linking group (Figure 9b), per-O-acetylated 5-aryl azide-N-acetylneruaminic acid (Ac₅-5-AAz-NeuNAc)¹⁹¹ and per-O-acetylated 9-aryl azide-N-acetylneruaminic acid (Ac₄-9-AAz-NeuNAc),¹⁹⁰ were synthesized and metabolically incorporated into cellular glycans. To demonstrate the utility of these probes, Paulson and co-workers used Ac₄-9-AAz-NeuNAc to discover that CD22, a sialic acid-binding glycoprotein, forms homomultimeric complexes on the surface of B cells.¹⁹⁰ Paulson and co-workers also used Ac₄-9-AAz-NeuNAc to identify the trans ligands of CD22.³⁷³ In this study, the authors incubated a CD22 fusion protein with B cells that were metabolically labeled with 9-AAz-NeuNAc and then used UV irradiation to stabilize CD22–glycoprotein complexes. Following immunoprecipitation of CD22 and quantitative proteomics, they identified 27 cell-surface glycoproteins. These results highlight the potential of using photo-cross-linking reagents to identify glycan-dependent protein interactions.

In the design of unnatural monosaccharides as photo-cross-linking reagents, the size and position of the photo-cross-linking functional group are important considerations. For example, large photo-cross-linkers can reduce the efficiency of metabolic incorporation and impact glycan-epitope recognition.^374,375 To address these issues, Kohler and co-workers synthesized unnatural monosaccharides containing a diazirine, which is a small photoactivatable cross-linking functional group.³⁷⁶ The authors introduced a diazirine into the N-acyl side chains of sialic acid and ManNAc, generating two unnatural monosaccharides (Figure 9c): per-O-acetylated 5-diazirine-N-acetylneruaminic acid (Ac₅-5-DAz-NeuNAc) and per-O -acetylated N-diazirine-acetylmannosamine (Ac₄ManNDAz).^377,378 Both Ac₅-5-DAz-NeuNAc and Ac₄ManNDAz were incorporated into cell-surface glycoproteins in the form of sialic acid and were used to study ganglioside–protein interactions.^379,380 Kohler and co-workers also used Ac₄ManNDAz to identify cell-surface ligands for cholera toxin on colonic epithelial cells, which led to the discovery of fucosylated glycoprotein ligands that might constitute the true ligands involved in cholera in humans.³⁸¹ To enable both cross-linking and enrichment, Chen and coworkers synthesized a bifunctional sialic acid derivative that contained an azide at the C9 position and a diazirine on the N-acyl side chain (Figure 9c): 5-diazirine-9-azido-N-acetylner-uaminic acid (5-DAz-9-Az-NeuNAc).³⁸² Using Daudi cells as a model system, the authors incorporated 5-DAz-9-Az-NeuNAc into cell-surface glycans, captured GBPs using UV irradiation, and enriched the complexes after conjugation to an affinity probe using the CuAAC reaction. To increase both the metabolic labeling and cross-linking efficiencies with these different sialic acid derivatives, the endogenous sialic acid residues can be enzymatically removed from cell-surface glycans.³⁸³

Beyond sialic acid and ManNAc derivatives, Kohler and coworkers also synthesized diazirine derivatives of GlcNAc (Figure 9d). The first analogue, per-O-acetylated N-diazirine-acetylglucosamine (Ac₄GlcNDAz), was not a compatible substrate for the GlcNAc salvage pathway.³⁸⁴ To overcome this barrier, Kohler and co-workers synthesized a peracetylated diazirine-functionalized GlcNAc precursor with an S-acetyl-2-thioethyl- (Ac-SATE-) protected phosphate group [Ac₃GlcNDAz-1-P(Ac-SATE)]. The authors also mutated the enzyme AGX1, which converts GlcNAc-1-P into UDP-GlcNAc. Using this combination of synthetic chemistry and genetic engineering, they demonstrated that HeLa cells cultured with Ac₃GlcNDAz-1-P(Ac-SATE) produce O-GlcNDAz-modified proteins.³⁸⁴ Subsequently, they employed structure-guided mutagenesis to enhance the efficiency of OGT with UDP-GlcNDAz as a substrate.³⁸³ In the future, it will be exciting to see diazirine derivatives of GlcNAc used to globally profile GBPs.

Although photo-cross-linkable monosaccharides have been used in relatively few studies and in limited-scale proteomics or glycoproteomics experiments, their ability to capture glycan-dependent interactions in a native cellular environment offers a powerful tool for glycobiologists. In the future, a wide range of glycan-dependent interactions could be identified by expanding the types of monosaccharides functionalized with cross-linkers. Additionally, by improving the computational tools required to analyze glycoproteomics data sets from cross-linked samples, it might become possible to identify the cross-linked sites within the GBP.

7. SUMMARY AND FUTURE OUTLOOK

Chemical tools have enabled tremendous growth in glycobiology, reducing experimental barriers that once restricted access to the study of protein glycosylation. In particular, chemical glycoproteomics has allowed the glycobiology community to enrich and identify glycoproteins and glycan structures associated with a number of biological processes. Continued technological advances are necessary to sustain this growth and exploration of glycoproteins and their biological relevance.

From the standpoint of basic research, it remains imperative that we strive toward a complete parts list for the human glycoproteome. We believe that it is a reasonable target that, within the next decade, the questions of which proteins are glycosylated at what sites with which structures, under specified conditions and with quantitative measurements, should be answerable. However, this state of knowledge will require accelerated advances in instrumentation; informatics; and the availability of reagents for binding, tagging, and enriching the various sectors of the glycoproteome. Thus, tool development remains central to the vision of a complete description of the human glycoproteome. As well, the tools that we now have on hand, powerful as they are, have not yet been democratized so that they can be wielded by nonexperts. For this reason, clinical translation of glycoproteomics is largely unrealized but constitutes an important and aspirational goal. Disease biomarkers are undoubtedly lurking in the glycoproteome, waiting to be discovered and exploited.

Important efforts that are well underway include technologies for the complete structural assignment of intact and complex glycopeptides, simultaneously resolving both the amino acid and glycan sequences at the proteome scale. New tandem MS technologies could allow techniques such as ETD to become more sensitive and routine. Ion-mobility MS offers an additional level of separation based on a molecule’s collisional cross section, which affects its rate of migration in the gas phase.³⁸⁵ This can be used to analyze complex glycopeptide mixtures, resolving identical peptides with different sites of glycosylation as well as isomeric glycan structures.^386–390 MS technologies are improving exponentially, and challenges in glycoproteomics should help direct that growth. Likewise, bioinformatics platforms capable of extracting glycosites and structures from complex MSⁿ data sets must keep up with new experimental technologies. (For more information, see the recent review by Hu et al.³⁹¹) Commercial analysis tools such as Byonic (Protein Metrics)³⁹² and SimGlycan (Premier Biosoft);³⁹³ several open source and academic analysis tools;^391,394 and database tools such as GlycoMod, Glyco-SuiteDB,³⁹⁵ and UniCarbKB^396,397 are already having an impact in this regard. As well, big data studies correlating disease-associated mutations and genetic variations with PTMs are emerging and should help frame studies of the glycoproteome.^398,399

Another exciting goal is the integration of spatial information into glycoproteomics data sets. Although MS is capable of identifying thousands of proteins in a given sample with high accuracy and reproducibility, their spatiotemporal relationships are typically lost in such experiments. Imaging techniques are typically employed for spatial and temporal analysis of molecules in cells and tissues and are widely employed for protein and, more recently, glycan analysis.^400,401 However, image analysis of specific protein glycoforms remains a significant challenge. Several laboratories have combined oligosaccharide metabolic labeling with protein labeling to image protein glycoforms using FRET-based microscopies.^402–406 In addition, protein-specific aptamers have been used to direct selective labeling to nearby glycans.⁴⁰⁷ These methods have yet to be scaled beyond a single protein target, but with multiplexable platforms such as MALDI imaging MS,^408,409 secondary-ion MS,⁴¹⁰ and mass cytometry (CyTOF),^411–413 there could be a rich future for the expansion of spatially resolved MS across the glycoproteome.

Biographies

Krishnan Palaniappan received his undergraduate degree in Chemistry from Carnegie Mellon University in 2005. He worked with Professor Bruce Armitage to study the behavior of DNA intercalating dyes. He then spent a year at the National Institutes of Health in Bethesda, MD, as a Postbaccalaureate fellow. Working in the lab of Dr. Kenneth Jacobson, he probed the binding sites of G-protein-coupled receptors for orthogonal activation using tailored ligands. In 2006, he attended graduate school at the University of California, Berkeley, where he worked jointly with Professors Carolyn Bertozzi and Matthew Francis. He was awarded his Ph.D. in Chemistry in 2012 for his research on the development of chemical tools and computational strategies for glycoproteomics and the development of viral capsid-based contrast agents for molecular imaging using xenon MRI. He is currently a Senior Scientist at Verily Life Sciences (formerly Google Life Sciences).

Carolyn Bertozzi is the Anne T. and Robert M. Bass Professor of Chemistry and Professor of Chemical & Systems Biology and Radiology (by courtesy) at Stanford University, as well as an Investigator of the Howard Hughes Medical Institute. She completed her undergraduate degree in Chemistry at Harvard University in 1988 and her Ph.D. in Chemistry from UC Berkeley in 1993. After completing postdoctoral work at UCSF in the field of cellular immunology, she joined the UC Berkeley faculty in 1996 and rose through the ranks to become the T. Z. and Armgard Chu Distinguished Professor of Chemistry and Professor of Cell and Molecular Biology. In June 2015, she joined the faculty at Stanford University coincident with the launch of Stanford’s ChEM-H institute.