Mass spectrometry for the identification and analysis of highly complex glycosylation of therapeutic or pathogenic proteins

Abstract

Introduction

Protein glycosylation influences characteristics such as folding, stability, protein interactions, and solubility. Therefore, glycan moieties of therapeutic proteins and proteins that are likely associated with disease pathogenesis should be analyzed in-depth, including glycan heterogeneity and modification sites. Recent advances in analytical methods and instrumentation have enabled comprehensive characterization of highly complex glycosylated proteins.

Area covered

The following aspects should be considered when analyzing glycosylated proteins: sample preparation, chromatographic separation, mass spectrometry (MS) and fragmentation methods, and bioinformatics, such as software solutions for data analyses. Notably, analysis of glycoproteins with heavily sialylated glycans or multiple glycosylation sites requires special considerations. Here, we discuss recent methodological advances in MS that provide detailed characterization of heterogeneous glycoproteins.

Expert opinion

As characterization of complex glycosylated proteins is still analytically challenging, the function or pathophysiological significance of these proteins is not fully understood. To reproducibly produce desired forms of therapeutic glycoproteins or to fully elucidate disease-specific patterns of protein glycosylation, a highly reproducible and robust analytical platform(s) should be established. In addition to advances in MS instrumentation, optimization of analytical and bioinformatics methods and utilization of glycoprotein/glycopeptide standards is desirable. Ultimately, we envision that an automated high-throughput MS analysis will provide additional power to clinical studies and precision medicine.

1. Introduction

Glycosylation is the most common post-translational modification (PTM) of proteins. The two major types of glycosylation, N-linked and O-linked, impact many functions, including protein folding and stability, protein interactions, and protein solubility [1]. Notably, changes in the N- and/or O-glycosylation patterns of various proteins have been reported in several diseases, such as cancer, infection, autoimmune diseases, diabetes, and chronic inflammatory diseases [1–3]. Glycosylation also influences therapeutic efficacies of protein drugs through changes of activity, pharmacokinetics, clearance, and immunogenicity [4]. To date, more than 100 proteins have been approved as therapeutics, and most of them are N- and/or O-glycosylated [5]. The glycosylation of biological products in the list of the US Food and Drug Administration’s Center for Drug Evaluation and Research are summarized in Table 1. With so many glycoproteins used as drugs and many more at different stages of development, it is critical to develop robust quantitative methods for in-depth analysis of protein glycosylation in terms of glycan structure and attachment sites. The same is true for pathogenic proteins to better understand their (patho)biological roles. While mass spectrometry (MS) approaches for quantitative proteomics have greatly advanced, the analytical workflow for highly glycosylated proteins still needs to be optimized for the evaluation of glycan structure and their site-localization. The aim of this review is to summarize recent technical developments and advancements in MS analysis toward understanding of the biological roles for therapeutic or pathogenic glycosylated proteins. We review general methodological advances for the analysis of glycosylated proteins and also provide specific examples of analytical methods, especially those using MS, that have enabled detailed analysis of highly complex glycosylation of therapeutic and pathogenic glycoproteins. These efforts will lead to a better understanding of the biological roles and specific function of glycoproteins.

Table 1.

Glycosylation of biological products in the list of the US Food and Drug Administration’s Center for Drug Evaluation and Research

Therapeutic proteins	Proprietary name	Definition	Production system	Class	Total glycosylation sites	Main target disease
Abatacept	Orencia	Anti-CTLA4	CHO	IgG1	6 N-linked 4 O-linked	Rheumatoid arthritis
Adalimumab	Humira	Anti-TNFα	CHO	IgG1	2 N-linked	Rheumatoid arthritis
Aflibercept	Eylea/ zaltrap	Anti-VEGF	CHO	IgG1	10 N-linked	Neovascular age-related macular degeneration, colon cancer
Agalsidase β	Fabrazyme	Agalsidase β	CHO		3 N-linked	Fabry disease
Alemtuzumab	Campath, Lemtrada	Anti-CD52	CHO	IgG1	2 N-linked	Chronic lymphocytic leukemia
Alglucosidase α	Myozyme	Alglucosidase α	CHO		6 N-linked	Pompe disease
Alirocumab	Praluent	Anti-PCSK9	CHO	IgG1	2 N-linked	Hyperlipidemia
Alteplase	Activase	Tissue plasminogen activator	CHO		3 N-glycan	Acute ischemic stroke
Asfotase α	Strensiq	Asfotase α	CHO	IgG1	12 N-glycan	Perinatal/infantile- and juvenile-onset hypophosphatasia
Avelumab	Bavencio	Anti-PD-L1	CHO	IgG1	2 N-linked	Non-small cell lung cancer
Basiliximab	Simulect	Anti-IL-2R α	Sp2/0	IgG1	2 N-linked	Kidney transplantation
Becaplermin	Regranex	PDGF-BB	yeast		2 N-linked	Wound care of diabetic foot
Belatacept	Nulojix	Anti-CTLA-4	CHO	IgG1	6 N-linked 4 O-linked	Kidney transplantation
Belimumab	Benlysta	Anti-B-cell activating factor	NS0	IgG1	2 N-linked	Systemic lupus erythematosus
Benralizumab	Fasenra	Anti-IL-5R	CHO	IgG1	2 N-linked (defucosylated)	Asthma
Bevacizumab	Avastin	Anti-VEGF-A	CHO	IgG1	2 N-linked	Colon cancer, lung cancer, etc
Bezlotoxumab	Zinplava	Anti-C. difficile toxins A and B	CHO	IgG1	2 N-linked	Clostridium difficile infections
Brodalumab	Siliq	Anti-IL-17R	CHO	IgG2	2 N-linked	Psoriasis
Burosumab-twza	Crysvita	Anti-FGF23	CHO	IgG1	2 N-linked	X-linked hypophosphatemia
Cemiplimab	Libtayo	Anti-PD-1	CHO	IgG4	2 N-linked	Cutaneous squamous cell carcinoma
Cerliponase α	Brineura	Human tripeptidyl peptidase-1	CHO		5 N-linked	TPP1 deficiency
Cetuximab	Erbitux	Anti-EGFR	Sp2/0	IgG1	4 N-linked	Metastatic colon cancer
Daclizumab	Zinbryta	Anti-IL-2R	NS0	IgG	2 N-linked	Multiple sclerosis
Daratumumab	Darzalex	Anti-CD38	CHO	IgG1	2 N-linked	Multiple myeloma
Darbepoetin α	Aranesp	Erythropoietin	CHO		5 N-linked 1 O-linked	Renal anemia
Denosumab	Prolia	Anti-RANKL	CHO	IgG2	2 N-linked	Osteoporosis
Dinutuximab	Unituxin	Anti-disialoganglioside	CHO	IgG1	2 N-linked	Neuroblastoma
Dulaglutide	Trulicity	GLP-1 analog	CHO	IgG4	2 N-linked	Diabetes
Dupilumab	Dupixent	Anti-IL-4R α	CHO	IgG4	2 N-linked	Atopic dermatitis, asthma
Durvalumab	Imfinzi	Anti-PD-L1	CHO	IgG1	2 N-linked	Urothelial carcinoma
Ecallantide	Kalbitor	Selective human plasma kallikrein inhibitor	Yeast cells		2 O-linked	C1-inhibitor deficiency or dysfunctional C1-inhibitor
Eculizumab	Soliris	Anti-complement protein C5	NS0	IgG2, 4	2 N-linked	Paroxysmal nocturnal hemoglobinuria, hemolytic uremic syndrome
Elosulfase α	Vimizim	N–acetylgalactosamine 6–sulfatase	CHO		4 N-linked	Morquio syndrome
Elotuzumab	Empliciti	Anti-SLAMF7	NS0	IgG1	2 N-linked	Multiple myeloma
Emapalumab	Gamifant	Anti-interferon γ	CHO	IgG1	2 N-linked	Hemophagocytic lymphhistocytosis
Emicizumab	Hemlibra	Activated Factor IX and Factor X	CHO	IgG4	4 N-linked	Haemophilia A
Epoetin α	Epogen	Erythropoietin	CHO		3 N-linked 1 O-linked	Renal anemia
Erenumab	Aimovig	Anti-calcitonin gene-related peptide receptor	CHO	IgG2	2 N-linked	Migraine
Etanercept	Enbrel Erelzi Eticovo	Anti-TNF-α	CHO	IgG1	6 N-linked 26 O-linked	Rheumatoid arthritis
Evolocumab	Repatha	Anti-PCSK9	CHO	IgG2	2 N-linked	Hyperlipidemia
Fremanezumab	Ajovy	Anti-CGRP	CHO	IgG2	2 N-lined	Migraine
Galcanezumab	Emgality	Anti-CGRP	CHO	IgG4	2 N-linked	Migraine
Galsulfase	Naglazyme	N-acetylgalactosamine 4-sulfatase	CHO		6 N-linked	Mucopolysaccharidosis VI
Gemtuzumab ozogamicin	Mylotarg	Anti-CD33	NS0	IgG4	2 N-linked	Acute myeloid leukemia
Golimumab	Simponi	Anti-TNF-α	Sp2/0	IgG1	2 N-linked	Rheumatoid arthritis
Guselkumab	Tremfya	Anti-IL-23	CHO	IgG1	2 N-linked	Psoriasis
Ibalizumab	Trogarzo	Anti-CD4 receptors	NS0	IgG4	2 N-linked	HIV infection
Ibritumomab tiuxetan	Zevalin	Anti-CD20	CHO	IgG1	2 N-linked	Non-Hodgkin lymphomas
Idursulfase	Elaprase	Human iduronate-2-sulfatase	Human cells		8 N-linked	Mucopolysaccharidosis II
Infliximab	Infliximab	Anti-TNF-α	CHO	IgG1	2 N-linked	Rheumatoid arthritis
Inotuzumab ozogamicin	Besponsa	Anti-CD22	CHO	IgG4	2 N-linked	Acute lymphoblastic leukemia
Interferon α2b	Intron A	Interferon α2b	CHO		1 O-linked	Hepatitis C virus infection
Interferon β1a	Rebif	Interferon β1a	CHO		1 N-linked	Multiple sclerosis
Ixekizumab	Taltz	Anti-IL-17A receptor	Human cells	IgG4	2 N-linked	Psoriasis
Lanadelumab	Takhzyro	Anti-plasma kallikrein	CHO	IgG1	2 N-linked	Hereditary angioedema
Laronidase	Aldurazyme	Alpha-L-iduronidase	CHO		6 N-linked	Mucopolysaccharidosis I
Mepolizumab	Nucala	Anti-IL-5	CHO	IgG1	2 N-linked	Eosinophilic asthma
Mogamulizumab	Poteligeo	Anti-CCR4	CHO	IgG1	2 N-linked (defucosylated)	Cutaneous T-cell lymphoma
Natalizumab	Tysabri	Anti-integrin α4-subunit	NS0	IgG4	2 N-linked	Multiple sclerosis
Necitumumab	Portrazza	Anti-EGFR	NS0	IgG1	2 N-linked	Non-small cell lung cancer
Nivolumab	Opdivo	Anti-PD-1	CHO	IgG4	2 N-linked	Non-small cell lung cancer
Obinutuzumab	Gazyva	Anti-CD20	CHO	IgG1	2 N-linked (defucosylated)	Chronic lymphocytic leukemia
Ocrelizumab	Ocrevus	Anti-CD20	CHO	IgG1	2 N-linked	Multiple sclerosis
Ofatumumab	Arzerra	anti-CD20	NS0	IgG1	2 N-linked	Chronic lymphocytic leukemia
Olaratumab	Lartruvo	Anti-PDGFR-α	NS0	IgG1	4 N-linked	Soft tissue sarcoma
Omalizumab	Xolair	Anti-IgE	CHO	IgG1	2 N-linked	Asthma
Palivizumab	Synagis	Anti-F protein of RSV	NS0	IgG1	2 N-linked	RSV infection
Panitumumab	Vectibix	Anti-EGFR	CHO	IgG2	2 N-linked	Colorectal cancer
Pembrolizumab	Keytruda	Anti-PD-1	CHO	IgG4	2 N-linked	Melanoma, lung cancer, etc
Pertuzumab	Perjeta	HER2	CHO	IgG1	2 N-linked	Breast cancer
Polatuzumab vedotin	Polivy	Anti-CD79b	CHO	IgG1	2 N-linked	Diffuse large B-cell lymphoma
Ramucirumab	Cyramza	Anti-VEGFR2/KDR	NS0	IgG1	2 N-linked	Gastric adenocarcinoma
Ravulizumab	Ultomiris	Anti-complement component 5	CHO	IgG2	2 N-linked	Paroxysmal nocturnal hemoglobinuria
Raxibacumab	raxibacumab	Anti-protective antigen component of B. anthracis toxin	NS0	IgG1	2 N-linked	Anthrax
Reslizumab	Cinqair	Anti-IL-5	NS0	IgG4	2 N-linked	Asthma
Rilonacept	Arcalyst	IL-1R and the IL-1R accessory protein	CHO	IgG1	2 N-linked	Cryopyrin-associated periodic syndromes
Risankizumab	Skyrizi	Anti-IL-23 p19	CHO	IgG1	2 N-linked	Plaque psoriasis
Rituximab	Rituxan	Anti-CD20	CHO	IgG1	2 N-linked	Non-Hodgkin Lymphomas
Romosozumab	Evenity	Anti-sclerostin	CHO	IgG2	2 N-linked	Postmenopausal women osteoporosis
Sargramostim	Leukine	Recombinant GM-CSF	Yeast cells		2 N-linked 2–4 O-linked	Neutropenia
Sarilumab	Kevzara	Anti-IL6-Rα	CHO	IgG1	2 N-linked	Rheumatoid arthritis
Sebelipase α	Kanuma	Human lysosomal acid lipase	Egg white		6 N-linked	Lysosomal acid lipase deficiency
Secukinumab	Cosentyx	Anti-IL-17A	CHO	IgG1	2 N-linked	Psoriasis
Siltuximab	Sylvant	Anti-IL6	CHO	IgG1	2 N-linked	Castleman’s disease
Tenecteplase		TNK-tPA	CHO		2 N-linked 1 O-linked fucose	Acute myocardial infarction
Tildrakizumab	Ilumetri	Anti-IL-23	CHO	IgG1	2 N-linked	Psoriasis
Tocilizumab	Actemra	Anti-IL6R	CHO	IgG1	2 N-linked	Rheumatoid arthritis
Trastuzumab	Herceptin	Anti-HER2		IgG1	2 N-linked	Breast cancer
Ustekinumab	Stelara	Anti-IL-12	Sp2/0	IgG1	2 N-linked	Psoriasis
Vedolizumab	Entyvio	Anti-α4β7 integrin	CHO	IgG1	2 N-linked	Ulcerative colitis
Vestronidase α	Mepsevii	Lysosomal β-glucuronidase	CHO		4 N-linked	Mucopolysaccharidosis VII

The list of biological products was generated from Purple book (https://www.fda.gov/drugs/therapeutic-biologics-applications-bla/purple-book-lists-licensed-biological-products-reference-product-exclusivity-and-biosimilarity-or) on December 2019. Glycosylation of each products was referred from Drugs. com (https://www.drugs.com/), DRUGBANK (https://www.drugbank.ca/), and EUROPEAN MEDICINES AGENCY (https://www.ema.europa.eu/en).

cytotoxic T-lymphocyte associated protein 4, CTLA4; chinese hamster ovary, CHO; tumor necrosis factor, TNF; vascular endothelial growth factor, VEGF; proprotein convertase subtilicin/kexin type 9, PCSK9; programmed death-ligand 1, PD-L1; interleukin-2 receptor, IL-2R; platelet-derived growth factor-BB, PDGF-BB; murine myeloma cells, Sp2/0; murine myeloma cells, NS0; interleukin-5 receptor, IL-5R; vascular endothelial growth factor A, VEGF-A; interleukin-17 receptor, IL-17R; fibroblast growth factor 23, FGF23; programmed cell death-1, PD-1; epidermal growth factor receptor, EGFR; receptor activator of NF-κB ligand, RANKL; glucagon-like peptide-1, GLP-1; interleukin-4 receptor, IL-4R; signaling lymphocyte activation marker Family member 7, SLAMF7; calcitonin gene-related peptide, CGRP; interleukin-23, IL-23; interleukin-17A, IL-17A; interleukin-5, IL-5; CC chemokine receptor 4, CCR4; platelet-derived growth factor receptor-α, PDGFR-α; fusion protein of respiratory syncytial virus, F protein of RSV; human epidermal growth factor receptor 2 protein, HER2; vascular endothelial growth factor receptor 2 /kinase insert domain-containing receptor, VEGFR2/KDR; interleukin-5, IL-5; interleukin-1 receptor, IL-1R; interleukin-23p 19; IL-23p 19; granulocyte macrophage colony-stimulating factor, GM-CSF; interleukin-6 receptor α, IL-6Rα; interleukin-17A, IL-17A; interleukin-6, IL-6; Tenecteplase-tissue plasminogen activator, TNK-tPA; interleukin-12, IL-12.

1.1. The structure of N-glycans and the corresponding glycosylation pathways

N-linked glycosylation includes the attachment of N-acetylglucosamine (GlcNAc) residue of a glycan precursor to the nitrogen atom of an Asn side chain by a β−1 N-linkage. There is a common core for all eukaryotic N-glycans that includes three mannose (Man) and two GlcNAc residues attached to Asn, Manα1–3(Manα1–6)Manβ1–4GlcNAcβ1–4GlcNAcβ1-Asn. Based on additional sugar residues that extend from the core glycan, N-glycans are classified into three types: oligomannose, hybrid, and complex glycans.

N-glycan diversity is created during the process of N-glycosylation. N-glycosylation starts from the formation of a glyco-lipid precursor. A branched carbohydrate structure consisting of glucose (Glc)-containing glycans, (Glc)₃(Man)₉GlcNAc₂, is attached to dolichol phosphate that is then “flipped” into the lumen of the endoplasmic reticulum (ER) (Figure 1a). An oligosaccharyltransferase adds the carbohydrate chain to the Asn residue of the Asn-X-Thr/Ser consensus sequence of the nascent protein. After further removal of the Glc residues in the ER, the quality-control step, the folded glycoprotein moves to the cis face of the Golgi apparatus for additional removal of Man by α-mannosidase I (α-Man I; MAN1A1, MAN1A2, MAN1C1) to form Man₅GlcNAc₂ [2, 3, 6]. Further modifications are performed by GlcNAc-transferase (GnT)-I to form GlcNAcMan₅GlcNAc₂. Subsequently, the majority of N-glycans are trimmed by α-mannosidase II (α-Man II; MAN2A1, MAN2A2) to form GlcNAcMan₃GlcNAc₂ in the medial-Golgi. Once both Man residues are removed, a second GlcNAc is added to the C-2 of the α1–6 Man in the N-glycan core by the action of GnT-II to yield the precursor for all biantennary complex N-glycans; GlcNAc₂Man₃GlcNAc₂. Additional branches can be added at C-4 of the core α1–3 Man and at C-6 of the core α1–6 Man by GnT-IV (MGAT4A, MGAT4B) and GnT-V (MGAT5, MGAT5B). The “bisecting” GlcNAc residue can be attached to the β-Man of the core by GnT-III (MGAT3). Hybrid N-glycans are formed if the GlcNAcMan₅GlcNAc₂ glycan produced by GnT-I is not digested by α-Man II (MAN2A1,MAN2A2) [6]. Further modifications are performed by α1–6-fucosyltransferase (FUT8), β1,4-galactosyltransferases (Gal-T), α2,3-sialyltransferase (α2,3 Sialyl-T) and α2,6-sialyltransferase (α2,6 Sialyl-T), which is distributed in the medial to trans face of Golgi apparatus [3].

Figure 1.

Biosynthesis of common N- and O-glycans.

a) N-glycan synthesis pathway. The mature glycan precursor attached to dolichol phosphate (Dol-P) is transferred to the Asn residue of the Asn-X-Ser/Thr peptide sequence in the endoplasmic reticulum (ER) by oligosaccharyltransferase (OST). After cleavage of three glucose (Glc) residues and one mannose (Man) residue, properly folded glycoprotein with oligomannose N-glycan (gray shadow with one asterisk*) is transferred to the Golgi apparatus. In the Golgi apparatus, the structure of GlcNAcMan₅GlcNAc₂ is generated by further mannose removal and addition of GlcNAc. By additional trimming by α-mannosidase Ⅱ (α-Man II) and addition of second GlcNAc by GlcNAc-transferase-II (GnT-II), the precursor of all complex N-glycans is generated. Further branching and capping of antennae is performed by some transferases to synthesize complex N-glycans (gray shadow with three asterisks***): GlcNAc-transferase III-V (GnT-III-V), α1–6-fucosyltransferase (FUT8), β1,4-galactosyltransferases (Gal-T), α2,3-sialyltransferase (α2,3, Sialyl-T) and α2,6-sialyltransferase (α2,6 Sialyl-T). Hybrid N-glycans are formed if the GlcNAcMan₅GlcNAc₂ glycan is not trimmed by α-mannosidase II and then extended by Gal-T and Sialyl-T (gray shadow with two asterisks**).

N-acetylglucosamine, GlcNAc; α- glucosidases, α-Glc; α-mannosidase I, α-Man I; GlcNAc-transferase I, GnT- I.

b) O-glycan synthesis pathway. O-GalNAc are added to the protein in α-linkage to Ser/Thr by a polypeptide GalNAc-transferase GALNT) in the Golgi apparatus. Four core structures of O-GalNAc glycans are generated by extension of glycans by different enzymes. Less extended O-GalNAc glycans, which are linked to Ser/Thr, are called Tn antigen, T antigen, and Sialyl-Tn antigen and are associated with several types of cancer.

N-acetylgalactosamine, GalNAc; galactose, Gal; N-acetylglucosamine, GlcNAc; N-acetylneuraminic acid, NeuAc; core 1 β1–3-galactosyltransferase 1, C1GALT1; molecular chaperone of C1GALT1, COSMC; core 2 β1–6-N-acetylglucosaminyltransferases 1, C2GnT-1, GCNT1; β−1,4-galactosyltransferase, B4GALT; α2–6-sialyltransferase, ST6GALNAC1; α2–3-sialyltransferases, ST3GAL1; β1–3-N-acetylglucosaminyltransferase 3, B3GNT3; core 3 β1–3 N-acetylglucosaminyltransferase 6, B3GNT6; M-type N-Acetylglucosaminyl transferase 3, GCNT3; β1,3-galactosyltransferase 5, B3GALT5.

1.2. O-glycosylation

O-glycosylation most often occurs on Ser and Thr residues, which are hydroxyl functional amino acids. The most common sugars attached to Ser/Thr are GlcNAc and N-acetylgalactosamine (GalNAc) in humans. Attachment of O-GlcNAc occurs in proteins in the nucleus, mitochondria, and cytoplasm; these proteins play important roles in regulating cellular processes, such as epigenetics, gene expression, translation, protein degradation, signal transduction, mitochondrial bioenergetics, the cell cycle, and protein localization [7]. O-GlcNAc is dynamically added and removed from proteins by O-GlcNAc transferase and O-GlcNAcase, respectively. O-GlcNAc modification is reported to act in a reciprocal manner to O-phosphate modification, and their respective addition and removal can affect protein structure, activity, and function [8].

Protein modifications with O-GalNAc are often represented by mucins as the prototype. Mucins are the class of glycoproteins carrying the greatest number of O-GalNAc glycans (mucin-type O-glycans). O-GalNAc glycan is usually extended to form one of four common structures (core 1 to 4). In tumor-associated mucin glycoforms, shorter glycans are enriched, such as a single unextended GalNAc linked to Ser/Thr (the Tn antigen), T antigen, and sialyl-Tn antigen. These glycoforms are thought to be involved in tumorigenesis and/or progression [9] (Figure 1b).

O-GalNAc glycans are added to proteins in α-linkage to Ser/Thr by the polypeptide GalNAc-transferase (GALNT) in the Golgi apparatus. There are 20 GALNT isoforms in humans, and they are differentially expressed in cells and tissues [10, 11]. Multiple core structures of the O-GalNAc glycan are generated by glycan extension by different enzymes. Core 1 β1–3-galactosyltransferase 1 (C1GALT1) adds galactose (Gal) to the GalNAc residue and generates the core 1 (Galβ1–3GalNAc-O-Ser/Thr) structure. Production of active C1GALT1 requires molecular chaperone C1GALT1C1 (Cosmc) [9]. Core 2 β1–6 N-acetylglucosaminyltransferases-1 (C2GnT-1; GCNT1) adds a GlcNAc β1–6 branch to the GalNAc residue of core 1 structure and forms core 2 O-GalNAc glycans. Core 3 O-GalNAc glycans are synthesized by core 3 β1–3 N-acetylglucosaminyltransferase 6 (C3GnT6; B3GNT6). Core 4 O-GalNAc glycans are synthesized from core 3 O-GalNAc by adding a branch using M-type N-acetylglucosaminyltransferase 3 (GCNT3) [9]. These core structure can be extended by attaching further Gal and sialic acid (Figure 1b).

2. Strategy for the analysis of protein glycosylation

In general, protein-glycosylation analysis can use one of the three common strategies, to address the usual heterogeneity of the glycans in glycoproteins (Figure 2).

Figure 2.

The overall strategy for the analysis of protein glycosylation. (Left) For intact glycoprotein analysis, a glycoprotein is directly subjected to high-resolution hybrid MS analysis after reduction of disulfide bonds. (Right) For glycopeptide analysis, the glycoprotein is digested with a protease(s) and the resultant glycopeptides are enriched prior to MS analysis. Glycans are released by enzymatic or chemical methods, either from glycopeptides or from glycoproteins. As shown in the lower panel, various useful technologies already exist and are rapidly advancing.

peptide N-glycosidase F, PNGase-F; Hydrophilic interaction chromatography, HILIC; capillary electrophoresis, CE; differential ion mobility, DMS; collision-induced dissociation, CID; electro transfer dissociation, ETD; Quadrupole time of flight, Q-TOF.

The first is to characterize intact glycoproteins or glycoprotein fragments after reduction of disulfide bonds by use of high-resolution MS (HRMS). Methodologies that allow for intact glycoprotein analysis have become a recent focus of technique development due to the desire to characterize therapeutic antibodies and other biosimilars that are glycosylated. Demonstrating reproducible distributions of glycan heterogeneity is not only desired but required for regulatory approval. This analysis is usually done on purifications of recombinantly expressed glycoproteins and thus there are usually ample amounts of protein for analysis by intact methods. Furthermore, native MS can be applied for the characterization of plasma glycoprotein microheterogeneity toward understanding of protein-drug interactions, pharmacokinetics of proteins, and glycosylation status changes of proteins in various diseases[12–14].

As intact glycoproteins can be directly subjected to MS analysis, in-depth sample preparation is not required in this approach. The MS characterization is also quite complicated by the extensive heterogeneity of the attached glycans that yield multiple glycoforms in combinatorial manner.

The second strategy is to analyze glycopeptides obtained by enzymatic proteolysis of glycoproteins (Figure 2), which overcomes the problems associated with analysis of intact glycoproteins. This analysis is beneficial because it can identify site-specific heterogeneity in glycoproteins. Qualitative and quantitative information for both N-glycopeptides and O-glycopeptides can be obtained. Typically, glycoproteins are reduced and alkylated (using dithiothreitol and iodoacetamide, and then enzymatically digested by a specific endoproteinase(s) (trypsin, lysyl endopeptidase (Lys-C), endoproteinase Glu-C (Glu-C), or endoproteinase AspN (AspN)), and the resultant glycopeptides are analyzed by MS. For highly sialylated glycopeptides, additional neuraminidase treatment simplifies the glycan structure and reduces the structural variety of attached glycans. Glycopeptides can be enriched prior to MS analysis, to addres the issues related to their ionization efficiency that is lower than that of other peptides. Lectin affinity chromatography is a commonly used tool to enrich for glycopeptides [15, 16]. Typically, broadly selective lectins such as Con A are used [15]. Alternatively, sequential or multi-lectin columns are used for the selective enrichment of glycopeptides [17]. Hydrophilic interaction chromatography (HILIC) is also available for glycopeptide enrichment. HILIC resins, such as cellulose, Sepharose CL4B, silica, and zwitterionic types have been used [18]. Recently, HILIC/ reversed-phase (RP) stop-and-go-extraction tip (Stage Tip) has been developed for enrichment in small sample amounts [19]. The MS analysis of enriched glycopeptides is commonly carried out with matrix assisted laser dissociation (MALDI) MS or liquid chromatography (LC)-tandem MS (MS/MS) [20, 21]. In particular, the development of high-resolution hybrid mass spectrometers has improved glycopeptide characterization [22] due to their ability to unambiguously distinguish between isobaric glycopeptide mixtures via accurate mass. Also, the use of several fragmentation methods that enable tandem MS sequencing (MS/MS) to provide glycan structural information as well as site specific information about glycosylation heterogeneity. These fragmentation techniques include as higher-energy collision dissociation (HCD) / collision-induced dissociation (CID), and electron transfer dissociation (ETD) [23, 24]. Recently, the combination of these fragmentation methods, e.g., ETD/HCD (EThcD) has been used for their superior fragmentation efficacy [25].

The last strategy is based on the analysis of released N- or O-glycans from glycoproteins (Figure 2). It is currently the most recognized approach for quantitative analysis of the complete, detailed glycan structure, including monosaccharide composition, sequence, branching, and linkages. However, in the case of glycoproteins with multiple sites of glycosylation, the context of site-specific heterogeneity is lost. So often these last two strategies are combined for complete glycoprotein analysis. For released glycan analysis, glycans are typically released through enzymatic or chemical methods and analyzed either by high performance LC (HPLC) and exoglycosidase digestion, or by MS using various MS/MS techniques [5] and more recently ion mobility technology [26]. For N-glycan analysis, first, peptide N-glycosidase F (PNGase-F) digestion is frequently conducted for their release. PNGase-F releases all asparagine-linked oligosaccharides, but it is not effective if they are α1–3-fucosylated, which is a common core-modification in plants [27]. In contrast, for the release of O-glycans, a general O-endoglycosidase is not available. The available endoglycosidase, O-glycanase, only cleaves the core 1 type O-glycan, without any other modification. Thus, releasing of O-glycans is widely conducted by chemical methods known as reductive β-elimination. β-elimination cleaves the glycosidic bond using a strong alkaline reaction, followed by reduction of the reducing terminus with sodium borohydride (NaBH₄). An alternative method is oxidative release of O- and N-glycans [28]. Finally, the released glycans with a reducing terminal are analyzed, most of which are derived by permethylation or fluorescent derivatization prior to analysis. Permethylation of released glycans has become more standardized in recent years. Permethylation neutralizes the negative charge of sialylated glycans and improves glycan tandem mass spectra with high structural information [29–31]. In contrast, the labeling of released glycans with fluorescent reagents is most useful for their specific and highly sensitive detection. 2-aminopridine (PA) and 2-aminobenzamide (2-AB) are commonly used fluorescent dyes for HPLC approaches [32, 33]. Recently, procainamide was shown to enhance electrospray ionization (ESI) efficiency and fluorescence intensity [34]. Because 2-aminobenzoic acid (2-AA) carries one negative charge, it is suitable for capillary electrophoresis (CE) and MS analysis. 8-aminopyrene-1,3,6-trisulfonic acid (APTS), or 8-amino-nephtalene-1,3,6-trisulfonate (ANTS) are also used for CE methods [35, 36]. Additionally, underivatized glycan alditols after β-elimination treatment are often analyzed by LC-MS on a porous graphite carbon column [37, 38]. As an alternative approach, the ion mobility technique has gained attention for its ability to distinguish between different isobaric glycans. This technology enables gas-phase separation of ions with identical m/z values but different compositional structure, linkage, or branching [39, 40]. Further improvements are needed in the future for facilitating glycosylation analysis and quantitative data output.

The final step of protein glycosylation analysis involves MS-output data processing. Due to the complexity and time needed for MS-output data interpretation, specialized bioinformatic tolls have been developed. The acquired MS data need to be compared with the knowledge databases for the assignment of glycan/glycopeptide structures. Several automated programs were developed for analysis of released glycan, such as GlycoWorkbench, SysBioWare, MultiGlycan, Glycan Builder, and GRITS Toolbox [41, 42]. For in-depth structural analysis including sites of glycan attachment, several groups have developed algorithms that have been summarized by Cao WQ et al [43]. For the glycopeptide analysis, commercially available software solutions, such as Byonic and SimGlycan are powerful tools for the rapid detailed analysis of complex N-and O-linked glycan structures from the LC-MS/MS data of glycopeptides [42]. As manual data validation is still needed due to high false-discovery rates of automated tools, several types of bioinformatics softwares have newly been developed, as summarized by Yu A et al [44]. However, each software should be validated in the specific samples and most of software can be applied to the analysis of either of N-glycoform or O-glycoform.

3. Examples of pathogenic glycoproteins and MS analytical approaches

Glycosylation is influenced by the activities of glycosyltransferases and glycosidases. In pathological conditions, such as cancer and inflammatory diseases, the activity and expression levels of different enzymes may vary. Hence, the surface-exposed structures and branching of glycans tend to change [45–48]. Among them, changes in immunoglobulin N-glycosylation in different disease states have been well reported. In this section, we summarize representative approaches for the analysis of N-glycosylation in IgG, O-glycosylation in IgA1 and MUC1, and N-glycosylation in the HIV-1 envelope glycoprotein (Env) and influenza hemagglutinin.

3.1. MS approaches for analysis of N-glycosylation of IgG in clinical samples

Human IgG has four subclasses and consists of two light chains (κ or λ) and two heavy chains (γ-chains); the latter have three constant regions (C_H1, C_H2, and C_H3) and one variable region (V_H). IgG can be subdivided into the Fab (antigen-binding fragment) and Fc (crystallizable fragment) regions. N-linked glycosylation occurs at Asn²⁹⁷ of the C_H2 domain and also on 15% to 25% of the IgG Fab portions [49, 50] (Figure 3a). The ligands of the Fc region include Fc receptors (FcγRI, FcγRII, FcγRIII), the C1q component of complement, and the neonatal Fc receptor. Other receptors may also bind IgG, either free, in an immune complex, or in aggregates. N-linked glycosylation of Asn²⁹⁷ of the C_H2 domain in the Fc region impacts binding to and activation of Fcγ receptors and of the C1q component of complement and IgG retention in the circulation [51]. N-linked glycosylation of the Fab region is relevant to the affinity and avidity of antibodies for antigens [52, 53] and also to antibody half-life [54].

Figure 3.

Schematic representation of IgG1 and IgA1/2 with the cleavage sites of different proteases.

a) Scheme of IgG1 structure is shown in the top left. The example of enzymatic digestion for separation of IgG Fab (antigen-binding fragment) and IgG Fc (crystallizable fragment) moieties are shown in the top right. Digestion of IgG by pepsin and papain occurs below and above the disulfide bond, respectively. Conversely, IdeS and IdeZ proteases digest IgG at a specific site between Gly-Gly of the IgG hinge region, below the disulfide bond. Novel workflow of site-specific N-glycan analysis of IgG, which was reported by Bondt et al. [49], is shown in the bottom of the figure. IgG was captured and digested by IdeS protease on beads. The F(ab’)₂ fragment was collected in the flow-through (FT), and the Fc portion was collected after elution by 100 mM HCl.

b) Schemes of IgA1 and IgA2 structure are shown in the top. Backbone peptide of IgA hinge region (HR) produced by trypsin is shown in the middle. O-glycosylation mainly occurs at the serine (S) and threonine (T) residues indicated in red. Digestion sites in the IgA1 HR by IgA-specific proteases (Clostridium ramosum AK183, Streptococcus pneumoniae TIGR4, Haemophilus influenzae HK50) are also shown. N-glycopeptides of IgA1/2 produced by trypsin digestion are shown at the bottom. N-glycan attachment sites are shown in red.

Closed circles represent N-glycans. Open circles represent O-glycans. constant region 1, CH1; constant region 2, CH2; constant region 3, CH3; heavy-chain variable region, VH; light-chain variable region, VL

The oligosaccharides present in the IgG Fc region are of the complex biantennary type and are comprised of a core heptasaccharide, i.e., GlcNAc₂Man₃-GlcNAc₂, forming a biantennary N-glycan with variable content of fucose, bisecting GlcNAc, Gal, and sialic acid. Due to the variable modifications of the core heptasaccharide, at least 36 different structurally unique oligosaccharide chains may be found at Asn²⁹⁷ [55]. Furthermore, due to the asymmetric glycosylation to each heavy chain, there can be more than 400 IgG glycoforms [56]. In healthy subjects, more than 90% of serum IgG contains core fucose and 50 to 70% are galactosylated. Furthermore, 10 to 20% of IgG is sialylated and a small proportion of IgG glycans (about 10–15%) contains a bisecting GlcNAc [49, 57, 58]. The composition of glycans tends to change with age [59–62], sex [61], and pregnancy [27, 63]. Alteration of IgG glycoforms have also been reported in diseases. Rheumatoid arthritis is among the diseases in which reduced galactosylation and sialylation of IgG has been observed [64–66]. Similar detects of IgG galactosylation have also been found in patients with systemic lupus erythematosus, Crohn’s disease, and other chronic and inflammatory diseases [3, 56, 67–70]. To assess IgG glycoforms, various analytical methods have been developed. The standard N-glycan analysis begins with IgG purification followed by analysis of released N-glycans and analysis of N-glycopeptides.

With regard to purification of IgG, the most popular technique is affinity-purification using immobilized Protein A [71, 72], Protein G [73], and anti-human IgG Fcγ antibodies [54]. For isolation of antigen-specific antibodies, affinity chromatography can be used with the specific antigen immobilized on a column [74]. IgG separation using SDS-PAGE, followed by in-gel digestion and N-glycan release has also been reported [75–77]. To analyze N-glycans in the Fab and Fc portions individually, several approaches have been reported to separate the two regions of IgG. Pepsin digests on the C-terminal side of disulfide bonding in IgG-heavy chains [78] (Figure 3a). By dialysis using a size-selective membrane, F(ab’)₂ fragments and Fc fragments are separated in the retained and flow-through fluids, respectively [73, 78]. Furthermore, papain which digests on the N-terminal side of disulfide bonding in IgG-heavy chains has been used to cleave IgG to Fc and Fab moieties [54, 71, 79] (Figure 3a). After digestion, Protein A columns binds the Fc fragments while Fab fragments are in flow-through [80]. Bondt et al. introduced a high-throughput workflow for separation of the Fc and F(ab’)₂ regions of IgG [49] (Figure 3a). Briefly, IgG is captured from serum or plasma using anti-IgG-Fc coupled beads, followed by digestion using IdeS or IdeZ proteases, which are cysteine proteases from Streptococcus pyogenes or Streptococcus equi subspecies zooepidemicus that digest antibodies at a specific site between Gly-Gly of the IgG hinge region. After digestion by IdeS protease, the F(ab’)₂ fragment is collected in the flow-through and the Fc portion in the eluant. To analyze N-glycopeptides, tryptic digests are applied to LC-MS/MS [49]. To analyze N-glycans released by PNGase-F, N-glycans can be enriched by various solid-phase extraction (SPE) methods, such as HILIC [49, 81], RP, and porous graphitic carbon (PGC) chromatography [82]. Permethylation of sialylated glycans can be applied before MALDI to avoid ionization biases and desialylation during ionization [83]. However, permethylation or derivatization of carboxyl groups, including methyl amidation [84] and methyl esterification [85], which enable differentiation between α2,6- and α2,3-sialylated glycans, require long derivatization in harsh conditions for a purified glycan sample. Recently, a rapid derivatization protocol with 1-ethyl-3-(3-(dimethylamino)propyl)carbodiimide hydrochloride (EDC) and 1-hydroxybenzotriazole monohydrate (HOBt) in ethanol has been developed [86]. This ethyl esterification reaction can be applied to a complex sample mixture before glycan purification, which enables high-throughput analysis. Furthermore, ethyl esterification of sialic acids enables mass-based differentiation between α2,6- and α2,3-sialylated glycans [87]. Ethyl esterification is a technique for the chemical modification of sialylated glycans for enhancing stability and avoiding the loss of sialic acid residues [88, 89].

3.2. MS approaches for O-glycosylation of IgA in clinical samples

The human IgA exists in two subclasses, IgA1 and IgA2. In human serum, IgA1 represents approximately 90% of total IgA. IgA1 has two N-glycosylation sites in the C_H2 region and in the tailpiece (Asn²⁶³ and Asn⁴⁵⁹), and nine potential O-glycosylation sites (serine (S) and threonine (T)) in the proline-rich hinge region (HR) (Figure 3b). In contrast to IgG, the impact of N-glycosylation patterns of IgA on binding to FcαRI is controversial [90–94]; however, it appears that the receptor glycosylation impacts IgA binding. On the other hand, abnormal O-glycosylation of IgA1 HR has been reported to be involved in the pathogenesis of IgA nephropathy (IgAN). In IgAN, serum levels of IgA1 with Gal-deficient (Gd) HR O-glycans are elevated, and these glycoforms are enriched in glomerular IgA1 immunodeposits [95]. The elevated serum levels of Gd-IgA1 have been confirmed by lectin-based enzyme-linked immunosorbent assay (ELISA) using Helix aspersa agglutinin, Vicia villosa, Helix pomatia, and Caragana arborescens which bind terminal, i.e., Gal-deficient, GalNAc [96–98]. Recently, mAbs specific for IgA1 with Gd O-glycans of IgA1 (Gd-IgA1) were developed, and their high reactivity with serum IgA1 from IgAN patients was demonstrated [99, 100]. However, lectin and antibody-based analysis are indirect approaches, and the strategy for identifying what kind of glycoform increases in whole IgA1 can be demonstrated only by detailed glycoform analysis using MS-based analytical strategies.

Recently, quantitative change in IgA1 HR O-glycan composition through therapeutic responses of IgAN was reported by using MALDI-TOF-MS and trypsin-digested purified IgA1, with glycopeptide enrichment by a hydrophilic affinity method using Sepharose CL4B, and desialylation by 2 M acetic acid [101]. MS strategies based on MALDI ionization allow high-throughput analysis. However, detection of sialylated glycopeptides is not reliable in the current commercially available MALDI-TOF systems. MALDI-Fourier transform ion cyclotron resonance (FTICR)-MS provides the advantage of an intermediate pressure in the MALDI source, which leads to a significant reduction in sialic acid fragmentation. Bondt et al. performed quantitative assessment of IgA1 N- and O-glycosylation in pregnant women with rheumatoid arthritis vs. healthy subjects by using MALDI-FTICR-MS; trypsin digestion of purified IgA and HILIC enrichment of glycopeptides was used [102, 103] (Figure 4a). In this study, glycopeptide enrichment by two-step microtip-cotton HILIC SPE enabled the site-specific N-glycosylation analysis. Furthermore, Plomp et al. [104] reported the strategy of IgA N- and O- glycosylation analysis using Nano LC-ESI-TOF-MS (/MS). With these methods, more detailed site-specific analysis of N-glycosylation was reported without glycopeptide enrichment. The detected glycopeptide sequences are shown in Figure 3b. The Asn⁴⁵⁹-containing glycopeptide with a truncated C-terminus was observed in both reports. This phenomenon is in accordance with previous reports [90].

Figure 4.

Comparison of two analytical work flow for IgA-glycopeptides.

a) High-throughput strategy for analysis of IgA N- and O- glycopeptides. This is the method reported by Bondt et al [102]. Purified IgA was digested by trypsin after reduction and alkylation. *To improve separation accuracy, two-step Hydrophilic interaction chromatography (HILIC) enrichment was performed, which enabled site-specific N-glycopeptide detection. The merit obtained by this method is high-throughput analysis. By using ultrahigh resolution matrix assisted laser dissociation (MALDI)-Fourier transform ion cyclotron resonance (FTICR)-MS, detection accuracy was improved and sialic acid fragmentation was reduced. The representative mass spectra after two-step HILIC enrichment were shown in the middle. This MS spectrum was obtained from ref 102. under Creative Commons Attribution 4.0 International License (https://creativecommons.org/licenses/by/4.0/).

O-glycopeptide, Q1; non-truncated, T1; truncated Asn340 containing glycopeptide, U1; hexose, H; N-acetylhexosamine, N; fucose, F; N-acetylneuraminic acid, S; sodium adduct, *; unidentified non-IgA-related glycopeptide contaminant, #

b) In-depth IgA O-glycopeptide analysis. Purified IgA1 was digested by neuraminidase and trypsin. By using an online nano liquid chromatography (LC) system, glycopeptide enrichment is not necessary before measurement by MS. Representative mass spectrum is shown in the middle. LC system coupled to MS enabled differentiation between isomeric forms. By shortening the hinge-region (HR) peptide with IgA-specific protease, detection of the attachment sites was demonstrated by Takahashi et al. using LC-electron transfer dissociation (ETD)-MS/MS [111].

Although the IgA1 HR O-glycoforms vary with respect to the number of monosaccharides attached, such as GalNAc, Gal, and sialic acid, the population of glycoforms can be determined by using trypsin-digested glycopeptide analysis with MALDI MS or LC-MS [95, 105–107]. Quantitative analysis of O-glycan microheterogeneity and attachment sites should be obtained for comprehensive understanding of the pathogenic forms of IgA in IgAN. IgA1 HR, rich in Ser, Thr, and Pro, contains clustered O-glycans and shows high resistance to proteases, including trypsin. Tryptic digests of IgA1 HR generate a 38-mer peptide with 9 potential sites for O-glycosylation. Thus, comprehensive analysis of IgA1 HR O-glycoforms is challenging due to large molecular mass and multiple glycosylation sites. To address the problem, HRMS was applied to characterize O-glycoform microheterogeneity [108, 109]. Determination of sites with attached aberrant O-glycans was achieved by activated ion electron capture dissociation (AI-ECD) FTICR MS/MS [108]. The AI-ECD FTICR MS/MS protocol was subsequently demonstrated using three distinct IgA-specific proteases (Clostridium ramosum strain AK183, Streptococcus pneumoniae strain TIGR4, Haemophilus influenzae strain HK50) to reduce the molecular masses of precursor ions to obtain sufficient fragmentation of c or z ions to assign all O-glycosylation sites [110] (Figure 3b, 4). An extracted ion chromatogram (XIC) of some IgA1 O-glycoforms showed multi-modal peaks, indicating structural isomers based on glycan attachment sites [111]. Combined quantification of IgA1 HR O-glycoforms and localization of O-glycan attachment sites, including structural isomers, quantifies the sites of O-glycan heterogeneity (Figure 4b) [111, 112]. To increase the throughput toward analysis of clinical samples from IgAN patients, an automated program for quantitative analysis of the HR O-glycopeptide profiles was developed. Furthermore, sites with Gd O-glycans were unambiguously identified by electron transfer/higher-energy collision dissociation EThcD MS/MS following a removal of galactosylated O-glycans [113]. This novel protocol enabled quantitative assignment of Gd sites. The most frequent Gd site was T²³⁶, followed by S²³⁰, T²³³, T²²⁸, and S²³². Future improvement of these analytical methods will enable the analysis of clinical samples and the understanding of pathogenic IgA1 HR O-glycoforms in IgAN.

3.3. Tumor-associated glycoforms of MUC-1 as a potential therapeutic target

It is well recognized that glycosylation patterns of proteins are altered in multiple human diseases, including many types of cancer, namely different types of adenocarcinoma such as ovarian cancer and breast cancer. Mucin 1 (MUC-1) is a cell membrane protein that contains a large extracellular densely glycosylated domain (mucin domain), consisting of tandem repeats of 20 highly conserved amino-acid residues. MUC-1 is polymorphic, and the number of tandem repeats may range from 25 to 125. Each repeat has five potential O-glycosylation sites, all of which can be O-glycosylated. MUC-1 expressed by normal mammary-gland epithelial cells is generally glycosylated with larger branched core 2 O-glycans often extended by polylactosamine and, to a lesser extent, capped core 1 O-glycans, whereas the truncated core 1 structures and terminal or sialylated GalNAc are absent. In contrast, breast-cancer cell lines produce MUC-1 with altered glycosylation, generally characterized by high-density glycosylation and reduction in the O-glycan chain lengths [114]. MUC-1 is overexpressed and aberrantly glycosylated in many types of adenocarcinomas [115–117]. The aberrant MUC-1 glycoforms can apparently induce immune responses, as many patients with adenocarcinoma have antibodies specific for these cancer-associated MUC-1 glycoforms (for review see [118, 119]). Some observations indicated that such antibodies may be protective [120, 121], making MUC-1 a possible therapeutic target for immunotherapy [119, 122–126]. Specifically, MUC-1 (glyco)peptides or MUC-1 variants with multiple tandem repeats have been tested in clinical trials for the induction of immune responses against cancer-associated MUC-1 glycoforms (for review see [127]).

Analysis of MUC-1 glycosylation has relied on MS, especially to obtain site-specific information [128–132]. In principle, individual MUC-1 repeats can be released by a proteolytic digest (e.g., trypsin) and various glycoforms, separated by chromatographic methods. The sites of glycan attachment can be determined by fragmentation methods, such as ECD [132]. Analysis of mucins, such as MUC-1 by MS techniques is one area where progress has been slow due to the complexities of multiple tandem repeats with clustered O-glycans. However, as MUC-1 has become a target for immunotherapy [127], the demand for detailed analysis of MUC-1 O-glycans will likely increase. Especially as it has been shown that glycosylation of MUC-1 is important for the generation of high-affinity therapeutic antibodies [133].

3.4. Virus glycoconjugates for vaccine development targeting HIV-1 and influenza infections

Many enveloped viruses, including HIV-1 and influenza virus, encode envelope glycoproteins. Some of these glycoproteins have important functions in virus biology. Moreover, some of the glycosylated motifs are recognized by the host’s immune system and, thus, can trigger a response against the glycosylated compound(s) of the virus(es). These glycoproteins are of great interest to vaccinologists who have been working on the development of vaccines against HIV-1 and influenza.

Recent progress in HIV-1 vaccine development has been driven by better understanding of the HIV-1 envelope (Env) glycoprotein and the impact its glycans play in immune responses and, conversely, virus immune escape and protection from immune recognition. Viral proteins are dependent upon the host cellular machinery for protein synthesis and any subsequent post-translational modifications, including glycosylation. The HIV-1 Env glycoprotein, a trimer consisting of gp120 and gp41 non-covalently associated subunits, has multiple sites of N-glycosylation. The densely glycosylated surface of the HIV-1 Env glycoprotein, often referred to as a “glycan shield”, is host-dependent and yet successful in evading immune responses. Longitudinal sampling of HIV-1-infected individuals revealed changes in the locations of clustered N-glycans due to the high mutation rate of HIV-1, creating an ever-shifting surface topology of this glycan shield. In spite of this evasion mechanism, some individuals with chronic HIV-1 infection develop broadly neutralizing antibodies that recognize multiple HIV strains. Several of the epitopes recognized by the broadly neutralizing antibodies have been mapped to the HIV-1 Env glycoprotein segments with highly conserved sites of N-glycosylation. These glycoprotein epitopes are the focus of several vaccine development projects for HIV-1 (for review see [134–139]).

The influenza virus also uses protein glycosylation as a key component of its host-cell entry and immune evasion. Specifically, the surface proteins of the viral particles, hemagglutinin and neuraminidase, play vital roles in infection. The number of N-glycans in the hemagglutinin globular domain changes as it circulates in the human population. Although the N-glycans on hemagglutinin are not as densely clustered as those found on HIV-1 Env, they play a role in immune evasion by masking antigenic sites from antibodies (for review see [140]). However, soluble C-type lectins, such as the mannose-binding lectin and surfactant protein, are known to recognize influenza hemagglutinin glycans and promote innate immune responses by macrophages and other immune mechanisms [141, 142]. Viral hemagglutinin protein binds sialic acid groups of cellular surface proteins to achieve viral attachment and entry. Human influenza viruses bind preferentially to α2,6-linked sialic acid, whereas avian influenza viruses use α2,3-linked sialic acid. Amino acids at key positions in the hemagglutinin receptor that affect binding specificity have been identified in the seasonal human and avian viruses [143]. The influenza neuraminidase is vital for viral entry into the respiratory tract to bypass the sialic acid-rich mucus proteins. Neuraminidase is also required for release of newly formed viral particles from the cell surface of host cells [144]. Thus, drugs that are structural mimics of sialic acid can inhibit the activity of influenza neuraminidase and represent a viable treatment option (for review see [145]).

As glycosylation of HIV-1 gp120-gp41 and influenza hemagglutinin varies and impacts immune recognition and immune responses, great efforts have been made to better delineate the glycosylation of these viral glycoproteins. These approaches started with assessment of overall glycosylation of viral glycoproteins [146–150] and now can include analyses of occupancy of individual potential glycosylation sites and determination of the category of glycans attached, i.e., high-mannose/hybrid vs. complex glycans, using recombinant glycoproteins and envelope glycoproteins isolated from virions [151], as well as detailed analysis of heterogeneity of the glycosylation sites of these glycoproteins [152–160]. Based on information stemming from these analyses and additional structural studies, models of HIV-1 Env trimers have been generated to assess the role of glycosylation [161–164]. It is hoped that combined approaches that include detailed characterization of glycosylation of viral glycoproteins will enable development of suitable and effective vaccine candidates that can elicit broadly protective humoral responses against HIV-1 as well as influenza and other enveloped viruses [165–172].

4. Examples of therapeutic proteins and MS analytical approaches

Many protein therapeutics have been approved for clinical use by the US FDA and other regulatory agencies (Table 1). In the early phase, protein therapeutics were purified from native source, such as β-glucocerebrosidase from human placenta and insulin from bovine and porcine pancreas [173]. Currently, most therapeutic glycoproteins are produced as recombinant products. There are many advantages of recombinant protein therapeutics. First, by transcription and translation of an exact human gene, specific activity can be expected without immunological rejection. Second, by expressing a specific protein in a gene-engineered system, such as bacteria, yeast, insect cells, mammalian cells, transgenic animals, or plants, the recombinant proteins are produced more efficiently and at a lower cost. Third, these processes reduce the risk of animal or human illness, such as those from viral or prion diseases. Lastly, recombinant proteins can be modified (e.g., glycosylation, phosphorylation, proteolytic cleavage) to improve function by selecting the production system or selecting a particular gene variant. Due to the increased development of recombinant monoclonal-antibody (mAb) therapeutics and the effect of glycosylation on their function and pharmacokinetics, many glyco-analytical approaches have been developed for the analysis of N-glycosylation of recombinant mAbs, including human and humanized IgG.

4.1. N-Glycosylation patterns of monoclonal antibodies and the approach for glycan analysis

Among protein therapeutics approved by FDA, there are many mAbs (Table 1). It is estimated that 30% of newly licensed drugs in the next decade will be based on antibodies and their derivatives [174]. Most therapeutic mAbs are of the IgG class and contain a glycosylation site in the Fc region of each heavy chain, such as the human IgG [175, 176]. There are also glycosylation sites in the Fab region in rare cases such as cetuximab at Asn⁸⁸ of the VH region [175–179].

The glycosylation patterns of mAbs mostly depend on producing cell lines, although other parameters of the fermentation process (e.g., glucose content, dissolved oxygen, bioreactor pH, sodium butyrate, ammonia content, CO₂ concentration, temperature, and production scale) can also impact the final glycan patterns to some extent [180]. From the viewpoint of immunogenicity, the type of cell line used to produce a mAb is important. Most approved mAbs are produced using Chinese hamster ovary (CHO) cells because of their ability to produce human-like glycosylation. Some are produced in murine myeloma cell lines (e.g., NS0 or SP2/0). Products produced using murine cell-lines contain non-human glycan moieties, such as N-glycolylneuraminic acid and α-Gal, which are immunogenic in humans. Indeed, cetuximab, which is produced in the SP2/0 cell-line, has been reported to induce anaphylaxis in some of the recipients [181].

N-glycosylation patterns of mAbs also impact their pharmacokinetic properties, bioavailability, and biological functions. The specific Fab glycans, such as those with terminal Gal, can bind to asialoglycoprotein receptor, whereas Fc high-mannose glycans can bind to mannose-binding lectin. These interactions may impact half-life of these proteins in the circulation [175, 176, 182, 183]. Glycosylation of the antibody in the Fc region can also alter the effector functions of the antibody. For example, IgG glycoforms with glycans decorated with terminal GlcNAc may activate the complement system though the lectin pathway, IgG with Gal and sialic acid may bind C1q, whereas other types of glycosylation may impact complement-dependent cellular cytotoxicity (CDC) [184]. Furthermore, afucosylation (lack of core fucose) activates antibody-dependent cellular cytotoxicity (ADCC) via increased binding of the IgG Fc portion to FcγRIIIa on the natural killer cells [185]. As IgG glycosylation of Fc and Fab is crucial for pharmacokinetics and antibody effector functions, close monitoring of these characteristics during manufacturing is required.

To fulfill this requirement, many different methods for glycosylation analysis have been developed. Methods for glycosylation analysis of mAbs are classified into three main strategies, as described in section 2 (Figure 2), 1) intact molecule, 2) glycopeptide, and 3) released glycan. For a top-down or middle-down approach, an intact IgG molecule after reduction of disulphide bonds (or after digestion with endoproteinase LysC, Pepsin, or IdeS for middle-down) is directly analyzed by LC-HR MS [186, 187].

In the glycopeptide approach, a therapeutic antibody is cleaved with proteases such as trypsin, LysC or pronase. In-depth analysis of IgG glycans in a site-specific manner can be achieved [186, 188, 189]. To overcome ion suppression of glycopeptide signals by peptides, enrichment of glycopeptides by HPLC or HILIC is required before MS assay. Reusch et al. tested a high-throughput workflow for monitoring Fc-IgG glycosylation using LC-ESI-MS analysis after IgG purification by Protein A in immobilized 96-well plates and extraction of glycopeptides by HILIC beads [190].

N-linked glycans on glycoproteins can be released by PNGase-F or by other endoglycosidases like Endo H, Endo F, or Endo S. Released glycans can be rapidly profiled by MALDI-TOF MS after glycan permethylation or sialic acid esterification [188, 191]. Unlabeled N-glycans can be analyzed quantitatively by high-performance anion-exchange chromatography with pulsed amperometric detection [192].

4.2. N-glycosylation analysis of intravenous immunoglobulin (IVIG) and Hexa-Fc, a biomimetic for IVIG

Intravenous immunoglobulin (IVIG) is prepared by pooling IgG antibodies from the sera of healthy human donors. Although it was initially used as an IgG-replacement therapy, it has also been used as therapy for autoimmune/inflammatory conditions, such as idiopathic thrombocytopenic purpura, Kawasaki disease, myasthenia gravis, Guillain–Barré syndrome, and various polyneuropathies [193].

IgG mediates pro- and anti-inflammatory activities through the engagement of its Fc fragment with distinct FcγRs [194]. Human and mouse FcγRs consist of several activating receptors and one inhibitory receptor, FcγRIIB. After IVIG therapy, FcγRIIB is upregulated on myeloid cells and the number of FcγRIIB-expressing myeloid cells is increased [195].

N-glycosylation of the IgG Fc region was found to affect IVIG anti-inflammatory function via upregulating expression of FcγRIIB in the effector cells. About a decade ago, both removal of the Asn²⁹⁷-linked sugar moiety and removal of only the terminal sialic acid residues from IgG in IVIG preparations showed a loss of the anti-inflammatory activity of IVIG [196]. Recently, dendritic cell-specific intercellular adhesion molecule-3-grabbing non-integrin (DC-SIGN)-related protein 1 (SIGNR1) and its human counterpart DC-SIGN were identified as glycosylation-specific receptors for the IgG Fc region with terminal sialic acid-rich N-glycan. Binding of sialic acid-rich IgG to SIGN-R1/DC-SIGN on DCs was found to induce secretion of mediators that increase expression of FcγRIIB on the effector cells [194, 197].

Fokkink et al. investigated the variation in IgG Fc N-glycosylation between different batches and brands of IVIG products using the LC-MS technique [198]. The rates of sialylation of Fc N-glycans (range, min-max) were 13.5–18.4% in IgG1 and 16.7–21.5% in IgG2/3. Insufficient sialylation of Fc N-glycans requires extremely high doses of IVIG to be administered to patients (2g / kg body weight) for sufficient clinical effect. For reduction of cost and adverse events caused by excessive IVIG administration, the development of an IVIG biomimetic was expected.

More recently, hexametric human Fc-fusion protein (hexa-Fc), a biomimetic substitute for IVIG, was developed. Mekhaiel et al. succeeded in manufacturing polymerized human IgG1 based Fc-fusions. The 18-amino-acid tailpiece from human IgM together with a five-amino-acid linker was added to the C-terminus of IgG Fc-fusions mutated with His³¹⁰ to Leu to achieve a polymeric form [199]. After forming multimers, the affinity to FcRs of hexa-Fc was enhanced compared to that of IgG1-Fc monomers [199].

With regards to the interaction between hexa-Fc, the higher affinity of hexa-Fc to soluble recombinant human DC-SIGN compared to that of IVIG suggested that alteration of N-glycosylation affected the high affinity of hexa-Fc and DC-SIGN [200]. The difference in N-glycosylation patterns between hexa-Fc and IVIG was investigated by released N-glycans analysis using MALDI-TOF-MS. Hexa-Fc was found to be rich in high mannose glycans in comparison with IVIG [200]. Although the affinity of hexa-Fc to DC-SIGN (K_D of 1.26 μM) was lower compared to HIV gp120 (K_D of 4.39 μM), a well-studied DC-SIGN ligand known to carry high amounts of N-linked high mannose oligosaccharides, they demonstrated that measured off-rate of hexa-Fc from DC-SIGN were relatively slow (K_off of 8.39 $\times$ 10⁻⁴s⁻¹) [200]. This suggests that once hexa-Fc is bound to DC-SIGN, the binding of the complex is stable.

4.3. Glycosylation analysis of Fc fusion protein therapeutics

The therapeutic and preventive use of Fc fusion protein therapeutics has recently increased for the treatment of various diseases because of their favorable characteristics, such as extended serum half-life and easy effective purification by protein-G/A affinity chromatography [201].

Etanercept (ETN) is a fusion protein consisting of two copies of the tumor necrosis factor-α receptor (TNFR) and the Fc portion of human IgG1. ETN is produced in CHO cells and consequently bears high level of glycosylation. There are four N- and 26 O-glycosylation sites in the dimeric TNFR domain, as well as two N-glycosylation sites in the IgG Fc portion (Figure 5) [202]. Recently, Wohlschlager et al. reported a new approach for unraveling the glycoform heterogeneity of ETN using high-resolution native MS combined with enzymatic digestion [202]. Intact ETN was observed in a charge-state distribution from m/z 4800 to 6500 corresponding to charges of 25+ to 20+. By deconvoluting the raw mass spectra of intact ETN in the most abundant charge state, they observed at least 70 distinguishable protein signals within mass range from 125 to 131 kDa. This deconvoluted spectra obtained by native MS of intact proteins has the potential to be a fingerprinting tool for assessment of batch-to-batch variability. By virtue of the large number of existing glycan combinations on ETN, simplification of its structure by applying glycosidases and/or proteases was required. To determine N-glycosylation of TNFR and Fc domains separately, ETN was digested by IdeS at the hinge region (Figure 5 b-1 and b-2). The TNFR domain underwent further digestion by sialidase and O-glycosidase to acquire simplified spectra. To pinpoint the exact N-glycan structures on Asn¹⁴⁹ and Asn¹⁷¹, which are the N-glycosylation sites of TNFR domain, bottom-up analysis was performed using digestion with AspN followed by HPLC-MS/MS analysis (Figure 5 b-3). By combining top-down and bottom-up analyses, MS data showed that the TNFR domain most likely has the core N-glycan at Asn¹⁴⁹ decorated with Gal₂(GlcNAc)₂Man₃(GlcNAc)₂ (A2G2) and with Gal₂(GlcNAc)₂Man₃(Fuc)(GlcNAc)₂ (A2G2F) at the Asn¹⁷¹ glycan. With regard to determination of N-glycosylation of the Fc domain, the native MS result showed concordant results with bottom-up analysis of tryptic digested glycopeptides and (GlcNAc)₂Man₃(Fuc)(GlcNAc)₂ (A2G0F) / A2G0F and A2G0F / (Gal)(GlcNAc)₂Man₃(Fuc)(GlcNAc)₂ (A2G1F) were assigned as the most prominent decorations of the glycan attached at Asn³¹⁷ (Figure 5 b-3). Assessment of O-glycoforms was performed via the native MS approach after PNGase-F digestion with or without sialidase digestion. With both PNGase-F and sialidase digestion (Figure 5 a-2), 14–23 core 1 glycan unites were assigned in the deconvoluted spectrum as well as 17–21 O-core variants with different numbers of sialic acid residues that were detected in the deconvoluted spectrum with only PNGase-F digestion (Figure 5 a-1). The typical ratio of N-acetylneuraminic acid (NeuAc) content per O-glycan core was 1.2–1.3. Twelve attachment sites were identified by Houel et al. using ETD (Figure 5 a-3)[203]. As seen from these reports, native MS analysis has a rapid fingerprinting tool to provide valuable information on true glycoform heterogeneity at the intact protein level. However, specific deglycosylation and digestion by IdeS are required to increase accuracy and to find out site specific glycoform in such a highly complicated glycosylation analysis. Furthermore, integration of middle-down and bottom-up proteomics data is necessary for in-depth characterization that would include modified sites.

Figure 5.

The workflow for Etanercept N- and O-glycosylation analysis.

a) The analysis workflow for O-glycosylation of Etanercept. a-1) By native MS analysis after peptide N-glycosidase F (PNGase-F) digestion, the number of O-glycan core variants with different numbers of sialic acids was detected, and the ratio of NeuAc residues to O-glycan core was elucidated. a-2) After PNGase-F and sialidase digestion, the number of O-glycans was detected by native MS. a-3) attachment sites of O-glycans were analyzed by electron transfer dissociation (ETD) fragmentation after digestion by trypsin and sialidase.

N-acetylneuraminic acid, NeuAc

b) The analysis workflow of N-glycosylation of Etanercept. b-1) By IdeS digestion, followed by sialidase and O-glycosidase digestion, the N-glycoform of the tumor necrosis factor-α receptor (TNFR) domain was elucidated using native MS. b-2) After IdeS digestion, the N-glycoform of the Fc domain was elucidated using native MS. b-3) To pinpoint the exact N-glycan structures on Asn¹⁴⁹ and Asn¹⁷¹, glycopeptides digested by AspN were analyzed by HPLC-MS/MS. The N-glycan structures of Asn³¹⁷ were elucidated by trypsin-digested glycopeptide analysis.

endoproteinase AspN, AspN

4.4. Glycoform profiling of erythropoietin and its analog, darbepoetin

Erythropoietin (EPO) is a glycoprotein hormone which regulates erythrocyte production. It is produced primarily by renal peritubular interstitial fibroblasts [204]. Recombinant human EPO (rhEPO) has been approved for the treatment of anemia resulting from a chronic kidney disease, chemotherapy, certain anti-retroviral treatments, and many other diseases. Human EPO consists of 165 amino acids and has three N-glycosylation sites at Asn^24, Asn³⁸, and Asn⁸³, and a single O-glycosylation site at Ser¹²⁶ [205]. The main N-glycan structures of rhEPOs are complex-type tetra-sialylated tetra-antennary glycans [206, 207]. The glycosylation of EPO, especially sialic acid-containing carbohydrate content, influences its biological activity and pharmacokinetics. Darbepoetin-α, which is a glycosylation analog of rhEPO, best describes this phenomenon. With five amino-acid substitutions (Ala³⁰, His³², Pro⁸⁷, Trp⁸⁸, Pro⁹⁰ to Asn³⁰, Thr³², Val⁸⁷, Asn⁸⁸, Thr⁹⁰), two additional N-glycosylation motifs were added into the EPO protein [208]. By this modification, Darbepoetin-α attains two additional N-glycans (five in total) with higher sialic acid content compared to rhEPO. Higher sialic acid content extends serum half-life as it reduces serum clearance, but also imparts lower affinity for EPO receptors on the surface of red-blood cell precursors. Despite the negative impact of higher sialic acid content on the biological activity, the extension of half-life overcomes the reduced receptor-binding activity and enhances in vivo efficacy [208, 209].

Jiang et al. introduced the procedure of site-specific qualitative and quantitative analysis of N-and O-glycoforms in rhEPO. As there is no trypsin cleavage site between Asn²⁴ and Asn³⁸ and the resulting peptide ²¹EAENITTGCAEHCSLNENITVPDTK⁴⁵ is too large for MS analysis, endoproteinase Glu-C digestion was combined with trypsin digestion. As a result, rhEPO was cleaved at Glu (E) / Lys (K) / Arg (R). After enrichment of the intact glycopeptides by ZIC-HILIC material, enriched glycopeptides were analyzed by LC-ESI-MS. Site assignments for glycans were mainly achieved by accurate mass detection of Y₁ (peptide + GlcNAc) or Y₀ (peptide) ions. Up to 74 intact glycopeptides representing four glycosylation sites were detected, and extracted ion chromatograms enabled relative quantification of site-specific glycoforms.

5. Conclusion

Protein structure and biological function are altered by the addition or a change of glycosylation, both in pathogenic and therapeutic proteins. Although new analytical methods and techniques are regularly introduced in the field of glycoproteomics, comprehensive analysis of glycoprofiling still remains a scientific challenge, as the methods for elucidating the structure of composite glycans and modification sites are just beginning to be standardized across the analytical community rather than being the analytical specialty of a relative few. This standardization is being driven by the biotechnology and pharmaceutical companies with interest in having their therapeutic candidates approved by regulatory agencies. Along with this standardization and push in methodology, the opportunity to analyze native populations of glycoproteins isolated from disease states is increasing as well. While general characterization of glycoproteins by releasing the glycans and analyzing proteolytic digests is quickly becoming standard, the ability to assess glycoprotein heterogeneity with intact or minimally cleaved glycoproteins has the potential to become the method of choice in the future. Quantitative methods for assessing glycoprotein heterogeneity is the area that still needs the greatest development. As these methods are being improved and more widely disseminated, there will be a greater need for the standardization of how glycoprotein analysis is to be reported and compared across preparations and laboratories. Glycoprotein/glycopeptide standards will need to be developed as references. We expect that innovative analytical strategies will not only discover glycoforms of improved protein therapeutics or the glycan pattern related to disease pathogenesis, but will also lead to better understanding of the biological mechanisms and functional significance of protein glycosylation. High-throughput analysis will be required for the translation of glycoproteomic analyses to clinical research and application in precision medicine.

6. Expert Opinion

The expression patterns of some glycosyltransferases and glycosidases, the enzymes that compose the glycosylation pathways, are often altered in pathological conditions, such as cancer and autoimmune diseases [45–48]. Identification of the specific glycoforms in a disease and clarifying their relationship with clinical prognosis could potentially lead to the development of new biomarkers with diagnostic and prognostic significance. Furthermore, detecting the antigenic glycoforms that cause disease informs development of novel therapeutic drugs and vaccines, e.g., MUC1-peptide vaccines for cancer [127]. The glycosylation patterns of therapeutic proteins are important for their pharmacokinetic properties, biological functions, and safety. The glycoengineering of therapeutic proteins is becoming important for improving their therapeutic effect. Furthermore, in the near future, many biosimilars will be developed, and their structure (including primary structures, high-order structures, and PTMs), function, and safety have to be monitored through the development process. The detailed analysis of glycosylation in the reference product and comparison with glycosylation in newly developed biosimilar should be reported. Furthermore, lot-to-lot variability assessment of products may also be required.

To reach these goals, high-throughput analyses with high resolution are required. For highly accurate analysis, including quantitation of glycoform heterogeneity and localization of glycan-modified sites, digestion protocols using various proteases and separation and enrichment techniques should be considered to achieve sufficient intensity of glycopeptide ions for quantification and/or fragmentation [49, 205]. Furthermore, a combination of various glycosidases can be applied to simplify the glycan structures and make it easier to analyze the glycoform and glycan attachment sites [110, 113]. Sample preparation is an important factor for high throughput analysis. Better separation of target glycoproteins and/or efficient enrichment of glycopeptides improves throughput [49, 190]. The top-down or middle-down approach is one of the most promising techniques for high-throughput analysis, but further improvement is needed for sensitivity and detailed analysis of attached glycans. The analytical software should be optimized for glycoproteins, which contain multiple N- and/or O-glycosylation sites. Recent advances of MS-based proteomic techniques provide the establishment of a workflow for unbiased quantitative analysis of proteins in crude samples [210]. Furthermore, this analytical platform enables the discovery and validation of biomarkers, simultaneously using shotgun proteomics in great depth among different cohorts [211]. We also note that the quality of sample biobanking should always be considered, as glycans on proteins may be affected by sampling and storage conditions. Thus, a robust and high-throughput analytical platform should be applied to clinical samples to define pathological forms of glycoproteins and develop and characterize new biomarkers to fully extend the impact of the emerging field of “glycomedicine” [1–3].

Despite significant progress in understanding pathogeneses of various diseases based on genomic approaches, complexity of some diseases poses with formidable challenges. The molecular changes of pathogenic proteins are derived not just from DNA alterations, but from protein expression and modification, with the latter including enzymes related to PTMs. Key issues, such as heterogeneity of disease phenotypes, are often not clearly delineated from genetic/genomic studies. Recently, a proteogenomic approach using integration of quantitative unbiased proteomic profiling and/or phosphoproteomic profiling with genomic information has been applied to cancer research and has revealed new molecular subtypes of cancers that were not previously identified by genome sequencing [212, 213]. Thus, a multi-layer approach (genomics, proteomics, and metabolomics) using clinical samples has become a powerful combination of tools to connect genomic analyses with clinical phenotypes. Furthermore, a glycoproteogenomic approach indicates that protein glycosylation patterns associated with genetic variants can classify individuals [214]. A recent study showed that serum levels of abnormally glycosylated IgA1 were associated with regulation of expression of a specific glycosyltransferase [215, 216]. Glycoproteogenomic analyses using MS techniques offer a tremendous potential to elucidate the pathogenesis of diseases related to aberrant protein glycosylation.

Article highlights.

• Glycans on therapeutic proteins or pathogenic proteins should be analyzed not only for glycoform profiles, but also for their site specificities.

• Glyco-engineering of therapeutic proteins leads to better clinical efficacy and less side effects.

• Glycosylation of therapeutic proteins should be monitored carefully, as glycans affect efficacy or toxicity.

• Advances in MS instrumentation and a series of analytical methods enable in-depth characterization of heavily glycosylated proteins, such as IgA.

• Development of high-throughput characterization of glycosylated proteins in clinical samples could be applicable for precision medicine.