Putting the pieces together: Mapping the O-glycoproteome

Abstract

Protein glycosylation is a diverse and omnipresent protein modification. Glycosylation provides glycoproteins with important structural and functional properties to facilitate critical biological processes. Despite the significance of protein glycosylation, the investigation of glycoproteome, especially O-linked glycoproteome, remains elusive due to the lack of a comprehensive methodology to conform with the diversity of O-linked glycoforms of O-linked glycoproteins. In recent years, mass spectrometry has become an indispensable tool for the characterization of O-linked glycosylated proteins across biological systems. We herein summarize the recent developments and insights of MS-based O-linked glycoproteomic technologies, databases and bioinformatic tools, with a focus on O-linked intact glycopeptide analysis.

Introduction

Glycoproteins are decorated extensively with the most omnipresent and exquisite modification – glycosylation, which takes up ~ 50% of the proteome[1, 2]. The abundance and diversity of glycosylation validate the indispensable roles of glycosylation in various physiological and pathological functions including, but not limited to, cellular signaling, protein folding, apoptosis regulation, inflammatory responses, immune escape, and cancer metastasis[3, 4]. In protein glycosylation, monosaccharides, the basic building blocks of glycans are linked by glycosidic bonds through various enzymatic actions of glycosidases and glycosyltransferases[5, 6]. They can be anchored to at least nine different amino acid side chains[7]. The oligosaccharide modified glycoproteins are N-linked and O-linked glycoproteins. The biosynthesis of N-linked glycosylation is initialized by the covalent attachment of an oligosaccharide unit (14 monosaccharide residues) to the amide group of asparagine (Asn), then the oligosaccharide unit undergoes a sequential trimming and addition of certain monosaccharides by glycosidases and glycosyltransferases in a specific manner. Moreover, the N-linked glycosylation has a universal core glycan structure (GlcNAc₂Man₃) on a consensus peptide sequence: Asn-X-Ser/Thr, where X is any amino acid except proline (Pro). In contrast, the O-linked glycosylation has neither consensus sequences nor universal core structures. O-linked glycosylation (Figure 1A) is built up from the first monosaccharide (including galactosamine/GalNAc, mannose/Man, galactose/Gal, fucose/Fuc, glucose/Glc, and glucosamine/GlcNAc) that is attached to the hydroxyl groups of the side chains of serine (Ser) and threonine (Thr). Based on these structures, either linear or branched oligosaccharide chain of various lengths could be further extended. In both N-linked and O-linked glycosylation, glycans can be further modified with chemical modifications such as phosphorylation, sulfation, and acetylation, and other monosaccharides including Fuc and Sia in α-linkages, Gal, GalNAc and GlcNAc in both α- and β-linkages, and sulfate[8].

Figure 1.

A) Core structures for mucin-type O-glycans;B) Enrichment strategies for O-linked glycipeptides:1) Metabolic labeling; 2) Hydrazide chemistry; 3) Lectin weak affinity chromatography (LWAC); 4) EXoO; 5) Chemoenzymatic strategy; 6) HILIC; 7) "SimpleCell" strategy.

As a result of the structural diversity that comes with the biosynthesis of O-linked glycosylation, to characterize O-linked glycoproteome, which is the assignment of specific O-linked glycans to protein sites in a proteome level is challenging. Mass spectrometry-based approaches have been predominantly applied for the intact O-linked glycoproteome characterization[9, 10]. However, apart from the analytical challenges that come with the inherent complexities of O-linked glycosylation on specific glycosylation sites in intact glycopeptides, glycopeptides are often low in abundance, the heterogeneity generates various sub-stoichiometric modifications to the glycopeptide that further decrease the quantities of a unique glycoform[11]. Moreover, the selection of glycopeptides for mass spectrometry analysis can be further competed by the majority of coexisting unmodified peptides[12*]. Coupled with the fact that glycosylation tends to lower the ionization efficiency[13], it is necessary to increase the robustness of glycopeptide analysis by glycopeptide enrichments. Additionally, MS fragmentation methods do not cleave glycans and peptide portion with the same efficiency. For instance, collision-induced dissociation (CID) generates mostly fragment ions from the glycan portion of a glycopeptide, and the cleavage preferably happens at peptide backbone of glycopeptide for electron-capture dissociation (ECD) and electron-transfer dissociation (ETD)[14], whereas higher-energy collisional dissociation (HCD) cleaves at both peptide backbone and glycan portion of intact glycopeptides[15, 16]. Hence, a hybrid fragmentation approach[17], such as electron transfer dissociation with HCD supplemental activation (EThcD)[18**] would be beneficial for the profiling of intact O-linked glycoproteome. At last, the limited availability of adequate databases and bioinformatic tools for O-linked glycoproteomic interpretation is also hindering the advancing in O-linked glycoproteomic research.

Current Advances in O-linked Glycopeptide Enrichment and Separation

Efficient selective enrichment and separation regimes are crucial for the glycoproteomic characterization of O-linked glycoproteome[19]. Currently, a selection of enrichment methodologies (Figure 1B) of specific glycopeptide subtypes has been adopted to this purpose, including hydrazide chemical tagging[20], metabolic labeling[9, 11, 21-23*], chemoenzymatic labeling[22, 24, 25**], lectin chromatography[26, 27], HILIC[28, 29] and “SimpleCell” technology with homogenized O-linked glycans[30-32**], as well as a site-specific chemoenzymatic extraction of O-linked glycopeptides by EXoO[24, 25**, 33**].

Lectin Weak Affinity Chromatography

Apart from the traditional HILIC method, in which hydrophilic materials are used to capture the hydrophilic glycans anchored on the glycoproteins/glycopeptides, lectin weak affinity chromatography (LWAC) is the most adopted enrichment regime for peptides with O-linked glycosylation[19]. Lectins is a type of non-enzymatic proteins that recognize and bind certain types of glycans linked to proteins[10], they also exhibit different specificities towards glycans as well as the surrounding amino acid sequences of peptide backbone[34]. Lectins display varying specificity range, some recognize a wide range of glycan structures, so they are suitable for the enrichment of larger portion of glycopeptides with different glycans, while some have a narrower specificity, so they can be used for the enrichment of a small subset of the O-linked glycopeptides[35, 36]. Based on these features of lectins, single[37], serial[38] or multi-lectin[39] affinity chromatography could be adopted for glycoprotein purification. However, the available lectins and their combinations are neither efficient nor sufficient enough to capture the entire glycoproteome[36]. To render the complexity down, in an approach called “SimpleCell”, researchers homogenized O-linked glycans by blocking the elongation. Combining lectin enrichment and electron transfer dissociation (ETD)/higher energy collisional dissociation (HCD)-based MS analysis, 600 O-linked glycoproteins were identified with ~3000 O-linked glycosylation sites on these engineered human cells[40].

Hydrazide Chemistry

Another popular enrichment method in the field of glycoproteomics is hydrazide chemistry with a rather high specificity. Once oxidized (by periodate or certain kind of enzyme), carbohydrate chains would be covalently conjugated to hydrazide beads. Target glycoproteins/glycopeptides could be selectively released from the beads using chemical or enzymatic approaches[41, 42]. In the study published by Zheng et al.[41], Tn antigen (O-GalNac) was oxidized specifically using galactose oxidase. The peptides with oxidized Tn antigen would then be captured by hydrazide beads and released by methoxylamine (CH₃ONH₂). After LC–MS/MS analysis, 96 glycoproteins with Tn antigen were identified in Jurkat cells[41]. In particular, sialyl-Tn antigen is characterized pervasively due to its pathological implications[3*]. Sialic acid residues are easily oxidized under mild periodate conditions, then oxidized sialic acid residues can be captured by hydrazide beads. Moreover, since the glycosidic bond between sialic acid residues and monosaccharides are less stringent, sialylated O-linked glycopeptides could be released by hydrolyzing the sialic acid residues-linked glycosidic bonds using acids such as formic acid[19].

Metabolic Labeling and Chemoenzymatic Labeling

Ever since it was first introduced, metabolic labeling has been widely utilized for the characterization of O-linked glycoproteins/glycopeptides. The basic regime introduced by the Bertozzi group[21] is to replace the natural GalNAc with GalNAz, on which the azido group could be modified with some affinity tags such as biotin. Thus, subsequent enrichment is made possible. Woo et al. [9, 11] reported another mass-independent platform–IsoTaG, for intact glycopeptides analysis based on metabolic labeling and ETD-MS/MS. The isotope recoding group of their affinity probes contains two bromide atoms–⁸⁰Br and ⁸²Br, which have equal natural abundance. IsoTaG uses an algorithm to search for MS1 spectrum of enriched O-glycopeptides with “1:2:1 type” peak intensity, then tandem MS analysis would be directed to these identified isotopically recorded species. IsoTaG stands out for its high confidence in the identifications of O-linked glycopeptides. Sialic acid residues are also of interest for metabolic labeling, since sialic acid residues are often found at the termini of various glycan structures and they are often associated with cancers[43]. Similar rationale is also adopted in chemoenzymatic labeling, where a chemically active group, e.g. an azide, acetone, or alkyne, is linked to the GlcNAc residues of O-linked glycopeptides under enzymatic catalysis[22, 23*]. These chemical active groups provide docking sites for enrichment tags, which could be captured by functional resin or beads through mechanisms like “click-chemistry”[44**].

Site-specific Chemoenzymatic Extraction (EXoO)

Recently, with the discovery of a novel O-protease – OpeRATOR™ (Genovis Inc), which specifically cleaves O-linked glycopeptides at the N-termini of O-linked glycan-anchored Ser/Thr to release site-specific O-linked glycopeptides with their conjugated O-linked glycans, a highly-specific solid-phase enrichment (EXoO) of O-linked intact glycopeptides is made possible[24, 25**, 33**]. This robust chemoenzymatic enrichment method, coupled with ETD-MS/MS and HCD-MS/MS, has been proven to characterize O-linked intact glycopeptides with rather high specificity and throughput. EXoO could map over 3,000 O-GalNAcylation sites for the first time in complex biological samples as well as cell lines in one single experiment[33**]. Apart from OpeRATOR™, another mucinase StcE[45**] was recently employed in the “bump-hole” study of engineered GalNAc-T[46*, 47**], and the fragmentation approach comparison study for glycopeptides[18**].

Mass Spectrometry Fragmentation

To characterize intact O-linked glycoproteome, we need detailed information on both the peptide backbone and its associated glycan. We will focus on multiple tandem MS (MS/MS) strategies that have been developed and applied in the characterization for intact glycopeptides, including CID, ETD, and HCD (a comparison is shown in Figure 2).

Figure 2.

Comparison of different fragmentation regimes.

CID

Collision-induced/activated dissociation (CID/CAD) is the most common and robust fragmentation method used for peptide analysis[48]. CID breaks the peptide (-CO-NH-) bonds, producing b (N-terminal fragment) and y (C-terminal fragment) ions of peptides, leaving positive charge on either. Meanwhile, carbohydrate Y series ions are also created since glycosidic bonds of glycopeptide are more fragile than the amide bonds. Thus, the glycan moiety structure would be revealed predominantly[10]. Due to this difference in cleavage efficiency, CID is often applied in conjunction with other fragmentation regime for intact glycopeptides characterization[49]. Alternatively, adding one more fragmentation stage (MS3) to selected Y ion generated by MS/MS could be adopted to generate b and y ions from a peptide backbone[50].

ETD

Electron transfer dissociation (ETD) is an alternative MS fragmentation method[17]. During ETD fragmentation, random fragmentation along the peptide backbone is induced, albeit from gas-phase reactions with either thermal electrons or fluoranthene, respectively[14]. Multiply protonated peptide receives an electron from a radical anion and cleaves at the N-C_α bond to produces c (N-terminal fragment) and z (C-terminal fragment) ions[51]. Unlike CID, labile post-translational modifications, such as glycosylation, are generally retained on the peptide backbone and are largely unaffected by the ETD fragmentation process, making ETD an effective fragmentation regime for identifications of O-linked glycopeptides and glycosylation sites. ETD can be used alone for glycopeptide characterization. However, due to the inefficient cleavage of glycosidic linkages by ETD, the information of the monosaccharide units is barely obtained through ETD, only the monoisotopic masses of oligosaccharides could be obtained. To tackle this issue, hybrid fragmentation regimes that provide supplemental energy have been introduced and gained popularity, with ETD followed by supplemental HCD (EThcD) being one of the most robust glycoproteomic methods[18**].

HCD

Higher-energy collisional dissociation (HCD)is a state-of-the-art dissociation approach that is used extensively for high mass accuracy peptide analysis. HCD fragmentation also produces b and y ions. Compared to CID, which is often limited by the “one-third" m/z cutoff rule, HCD has higher trap efficiency in the low m/z range due to the fact that HCD fragmentation is performed in a dedicated octopole collision cell at the far end of the “C-trap”[52]. Moreover, small diagnostic oxonium ions of monosaccharaides could also be detected in HCD[53-55]. In addition, HCD could dissociate the peptide backbone efficiently as well as the attached labile glycans[55-57], HCD is therefore a desirable fragmentation regime for protein and glycopeptide identification and quantification. However, there’s tradeoff between proteome coverage and mass accuracy for HCD. In many cases, HCD is often combined with CID and/or ETD for complimentary fragmentation[55, 58, 59], which in general provides greater confidence in the identification of glycopeptides and glycosylation sites than studies relying on a single fragmentation regime[17].

Bioinformatics

The heterogeneity of O-linked glycoproteome has been challenging to conform with. A protein contains 4 glycosylation sites with 10 different glycans could yield 10,000 possible glycoforms, which at some point is simply impossible to incorporate into the database searching algorithms. Moreover, the glycans attached to the same protein vary in different cells, tissues, organisms and physiological states[60**]. The same glycan structure on different proteins may also exhibit different functions. Therefore, the interpretation with regard to the glycosylation should be comprehensive, integrating the glycosylation sites, the site-specific glycan structures, the protein scaffolds, and the co-expression of related glycogenes.

With the agglomerative efforts by scientists of all regimes, tremendous improvements in the analytical methods including high-performance separation strategies and mass spectrometry-based analysis have propelled the studies on glycosylation. Consequently, the qualitative and quantitative data of glycans, glycosites, glycopeptides, and glycoproteins have increased dramatically. Along with the increase in data availability, corresponding databases and bioinformatic tools to store, retrieve, integrate, and interpret these data have also grown substantially.

Databases for O-linked glycosylation

The glycan takes up a huge part of our O-linked intact glycopeptide analysis. Various glycan databases have been established ever since the first database in glycobiology – CarbBank[61]. CFG Glycan Structure Database[62] (http://www.functionalglycomics.org/glycomics/molecule/jsp/carbohydrate/carbMoleculeHome.jsp), Glycan Mass Spectral DataBase by JCGGDB[63] (https://jcggdb.jp/rcmg/glycodb/Ms_ResultSearc), UniCarb-DB[64] (https://unicarb-db.expasy.org/), GLYCOSCIENCE.de[65] (http://www.glycosciences.de/), GlyTouCan[66](https://glytoucan.org/) and Carbohydrate Structure DataBase[67] (http://csdb.glycoscience.ru/database/) are integrative databases of note, providing structural, chemical, mass spectrometry, origin and glycotransferase information of glycans.

Compared to N-linked glycosylation, there are relatively less data available for O-linked glycosylation due to the aforementioned challenges. Limited resources for O-GalNAc and O-GlcNAc proteins are made available through GlycoDomain Viewer[68] (https://glycodomain.glycomics.ku.dk/) and Glyco-DIA[69*] (https://github.com/CCGMS/Glyco-DIA) based on the “SimpleCell” technology, as well as NetOGlyc[70] (www.cbs.dtu.dk/services/NetOGlyc/), and YinOYang 1.2[71] (http://www.cbs.dtu.dk/services/YinOYang/).

Data Analysis Software for O-linked glycosylation

The recent published O-Pair Search[72**] (https://github.com/smith-chem-wisc/MetaMorpheus), utilized an ion-indexed open modification search strategy, to assign O-linked glycopeptides to paired HCD and ETD spectra. O-Pair Search is reported to be able to identify more O-glycopeptides compared to the most widely used commercial glycopeptide search tool, Byonic[73], while reducing search times significantly by more than 2,000-fold[72**]. Moreover, O-Pair Search also accepts user-defined glycan databases, potentiating searches other than O-GalNAcylation. Another mode within MSFragger (http://msfragger.nesvilab.org) —MSFragger-Glyco[74**], developed by Polasky et al., is also based on the ion-indexed search strategy. MSFragger-Glyco tremendously improved the identification of labile glycan spectra and the annotation of glycopeptide spectrum matches(glycoPSMs). For both O-Pair Search and MSFragger-Glyco, more glycopeptide assignments are achieved in less time compared to traditional search algorithms. Another newly developed searching tool for intact O-glycopeptides – AOGP[75*], is found to exhibit superior performance in identifying intact O-glycopeptides of single proteins.

Concluding Remarks and Future Perspectives

The field of O-linked glycoproteome study, especially the study of O-linked intact glycopeptides, is growing substantially over the recent years. Advancements in sample preparation and separation, instrumentation, specialized algorithms and bioinformatic tools, brought on many in-depth discoveries. However, the identified O-linked glycopeptides only comprise a small part of the entire O-linked glycoproteome. Challenges remain for O-linked intact glycopeptide analysis.

Although multiple enrichment methodologies have been made possible, all of them fall short in term of comprehensiveness or specificity. Affinity-based and labeling-based enrichment methodologies are limited to only a subset of glycan structures, while hydrophilic interaction-based methodologies cannot distinguish glycopeptides with other hydrophilic non-glycopeptides. To address this problem, a hybrid or multiplexed enrichment methodology could be designed to improve the coverage and throughput. In mass spectrometry-based analysis, the annotations of glycan structures often rely on oxonium ions and neutral loss. Since some monosaccharides have the same mass, it would be quite troublesome to pinpoint these monosaccharides and distinguish structure isomers[10]. Additionally, a hybrid fragmentation regime based on current available technologies could also benefit our analysis of O-linked intact glycopeptides[17], providing us with more details with regard to glycan structures and peptide backbones. Last but not least, a lag in the development for data integration and interpretation is also hindering the studies on O-linked glycoproteome. Future data analysis software hopefully will include more complete lists of glycospecific fragments, along with the ability to incorporate spectra generated from different types of fragmentation regimes. In recent years, we believe these challenges would be resolved with the rapid advancing of our technology.