Post-Translationally Modified Proteins: Glycosylation and Disulfide Bond Formation

Abstract

Even though most medicines have historically been small molecules, many newly approved drugs over the last two decades have been derived from proteins. For the past few years, protein therapeutics have been enjoying the fastest growth within the global pharmaceutical industry.

These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.

4.1 Introduction

Even though most medicines have historically been small molecules, many newly approved drugs over the last two decades have been derived from proteins. For the past few years, protein therapeutics have been enjoying the fastest growth within the global pharmaceutical industry. Protein-based therapeutics, such as insulin, interferons, monoclonal antibodies (mAb), growth hormones, erythropoietins, blood-clotting factors, colony-stimulating factors (CSFs), plasminogen activators, and reproductive hormones, play a significant role in the treatment of many major diseases, and protein therapies have revolutionized the methodology followed by drugs. These therapies exhibit high efficiency due to their targeted approach, which avoids side effects on healthy organs to a great extent. In recent years, the number of protein-based pharmaceuticals reaching the marketplace has increased exponentially, and they provide innovative as well as effective therapies for several chronic diseases which were previously not responsive to treatment. The global market for biologics or biotechnology therapeutics is one of the most prolific and fastest growing markets in the world, representing at least 24 and 22 % of all new chemical entities approved by the US and EU regulatory authorities, respectively [1]. Sales of biotech products in US showed an annual growth rate of 20 % between 2001 and 2006 compared with 6–8 % in the pharmaceutical market [2], and it is expected to grow at annual growth rate of around 13 % during the next three years (2012–2015), with the introduction of new protein therapeutics and enhanced investments contributing to this booming growth of this industry.

For protein therapeutics to be effective, they must be produced in biologically active forms, which require proper folding, and post-translational modifications (PTMs) with the extent of PTMs depending on the nature of the “host” cell and the conditions of the fermentation and recovery processes. Even though only a few bio-pharmaceutical proteins such as albumin (Recombumin) and insulin (Humulin N and Lispro) undergo simple modifications such that they can be manufactured using yeast or bacteria [3], most of the production platforms used to produce biopharmaceuticals comprise mammalian cells that have the ability to perform complex PTMs. The most prevalent modifications include variable glycosylation, formation of disulfide bonds, cysteine (C) and methionine (M) oxidation, phosphorylation, misfolding and aggregation, deamidation of asparagine (N) and glutamine (Q), and proteolysis at the C- and N-termini. Even though the presence of PTMs is often required for normal biological function or tissue disposition of the protein, in many cases, the role of the modification is as of yet unknown. Therefore, detailed characterization of these modifications is extremely important, because they may alter physical and chemical properties, folding, conformation distribution, stability, activity, which in turn may affect cellular processes, in which the protein is involved [4–6]. Examples of the latter can be regulation of signal transduction and a wide variety of cellular events such as growth, metabolism, proliferation and differentiation in case of protein phosphorylation [7–9], targeting, cell–matrix interaction, as well as pharmacokinetic and pharmacodynamic behavior in case of glycosylation [10, 11]. Therefore, a thorough verification of the protein’s (amino acid) sequence, assessment of the purity and impurities in a recombinant protein drug product along with a detailed characterization of the existing PTMs, is a regulatory requirement prior to its approval for clinical use [12].

Full structural characterization of the existing PTMs in a recombinant protein often poses a considerable analytical challenge owing to their inherent complexity. The presence of PTMs often complicates or even prevents the use of classical tools for protein sequence analysis (e.g., automated Edman degradation). Moreover, the presence of lipid or carbohydrate covalent attachments on proteins can dramatically decrease the accuracy of the molecular weight (M_r) measurement when using sedimentation velocity, gel permeation, or SDS-PAGE analysis. Separation techniques such as high-performance liquid chromatography (HPLC) or capillary electrophoresis (CE) combined with a variety of mass spectrometry (MS) techniques are commonly employed for the profiling and quantitation of PTMs present in recombinant therapeutic proteins. The development of electrospray ionization (ESI) [13, 14] MS coupled with online liquid chromatographic (LC–MS) or electrophoretic separation (CE-MS) [15, 16] and matrix-assisted laser desorption/ionization (MALDI) [17, 18] has established MS as the technology of choice for protein mapping, localization, structure identification, and quantification of existing PTMs [19, 20]. Several MS-based approaches have been developed employing tailored tandem MS scanning methods diagnostic for specific PTMs, such as monitoring precursor/product-ion transitions and neutral loss scan [21–23].

Recently, online LC–MS combined with collision-induced dissociation (CID) and electron-capture dissociation (ECD) [24] or electron-transfer dissociation (ETD) [25] fragmentation has been used to elucidate disulfide linkages and site-specific glycosylation in recombinant therapeutic proteins and glycoproteins [26, 27]. Similarly, MS-based approaches can be employed in the production of a recombinant therapeutic protein in order to ensure the purity, the production yield, and the absence of chemical degradation and/or aggregation products in the protein formulations for clinical and eventually commercial use.

In this chapter, we discuss MS-based methodologies that are employed to detect, identify, and characterize two of the most prevalent PTMs in the production of therapeutic recombinant proteins, glycosylation and disulfide bond formation. These MS-based approaches discussed here are representative of those used for the comprehensive characterization and quantitation of other PTMs encountered in recombinant proteins intended for therapeutic use in humans.

4.2 Glycosylation

Glycosylation process, that is, the covalent attachment of oligosaccharide chains on the protein backbone, is considered as the most important and common PTM of proteins. It is estimated that over 70 % of all human proteins are glycosylated [28] and 90 % of protein therapeutics are glycosylated [10]. The carbohydrate moieties of glycoproteins (glycans) can modulate the biological functions of a glycoprotein such as circulation, cell-to-cell interactions, receptor binding, molecular and immune recognition, which in turn affect intracellular signaling, fertilization, embryonic development, immune defense, recognition of hormones, cell adhesion, and pathogenicity [4]. In addition, glycan-chain modification can significantly impact their physicochemical properties such as protein folding, solubility, stability, aggregation, and susceptibility to proteolysis [29]. Finally, carbohydrate modifications can also considerably alter protein conformation, which may consequently modulate the functional activity of the protein, especially in its interactions with other proteins or ligands. It has been established that altered glycosylation or variation of a protein’s glycosylation pattern is associated with numerous diseases and disorders [30–32]. Therefore, detailed structural studies of the glycosylation and its inherent heterogeneity are also potentially vital toward understanding their function in complex physiopathological processes and establishing glycan profile changes between healthy and disease states [33, 34]. The latter has increased the potential of using glycan biomarkers for the diagnosis of several diseases [35], as well as for the design of new therapeutics [10, 36, 37]. Moreover, carbohydrate modification can be used toward the production of “custom-made” glycoproteins tailored, such as glycoproteins with defined homogeneous glycosylation structure, for specific therapeutic use [38].

Therefore, complete structural analysis of a glycoprotein end product will involve not only the determination of the primary peptide sequence, but also detailed analysis of the glycan structures including information on the individual glycosylation sites, the glycosylation patterns, and the structure elucidation of the attached carbohydrates (glycoproteome) [39–43]. As it has become obvious that many of the changes associated with disease and differentiation are due to the glycans attached to proteins (glycome), a thorough understanding of these glycan structures will be invaluable for gaining insight into their involvement in disease mechanisms and the potential for novel therapeutic interventions [44]. Characterizing the glycoproteome, however, is a challenging and daunting task because the structural heterogeneity of these glycans is vast, necessitating the development of highly sensitive and efficient analytical methods for detection, separation, and structural investigation of glycoproteins.

4.2.1 Intact Glycoprotein Analysis by Mass Spectrometry

An important preliminary step in the quality control and structure characterization of a therapeutic recombinant protein is the M_r determination of the protein product. On the intact glycoprotein level, non-spectrometric techniques such as SDS-PAGE, lectin affinity chromatography (LAC), isoelectric focusing (also in a capillary), or capillary zone electrophoresis (CZE) are generally used. In case of the two-dimensional (2D) gel electrophoresis separation of glycoproteins, characteristic spots reflecting their different isoelectric points and M_r of different glycoforms can be seen. The subsequent detection of the glycosylation pattern of the electroblotted glycoproteins may be performed by LAC [45, 46], where carbohydrate-specific lectins can be used to probe distinct oligosaccharide structures (motifs). In addition, this affinity purification can be employed as an enrichment method for the glycosylated peptides and proteins (see Sect. 4.2.2.2). Nevertheless, the low solubility of the membrane glycoproteins, resulting in their poor detection, is a significant drawback of the 2D gel electrophoresis approach. An alternative method of higher resolving potential is CZE or CE, where the various glycoforms are detected even though no information on the nature of the attached glycans is revealed [47]. These electrophoretic methods have been successfully used in the separation of sialic acid isoforms of endogenous and recombinant glycoproteins, and they have proved their usefulness in clinical diagnosis and product quality assessment [48].

In the late 1980s, the incorporation of ESI and MALDI MS, along with advances in electrophoretic separations and high-resolution MS, has provided a powerful analytical tool for the analysis and even quantitation of the intact individual glycoforms in glycoproteins [15]. ESI and MALDI MS are the premier methods of choice for M_r measurement and the ensuing protein mapping. In case of ESI MS analysis of therapeutic proteins, spraying of an aqueous protein solution at μL/min or nL/min flow rates generates multiply protonated signals with reduced mass-to-charge (m/z) ratios, thus making them readily detected by typical mass analyzers with a mass range up to 2,500 Da. This is demonstrated in the ESI mass spectrum of human recombinant interferon α-2b (INTRON A) (Fig. 4.1), which is used in the treatment of certain viral infections, including chronic hepatitis B, C, and D, malignant melanoma, follicular lymphoma, Kaposi’s sarcoma caused by AIDS, and infections caused by human papillomavirus (HPV). The ESΙ mass spectrum exhibited a bell-shaped distribution of multiply charged ions ranging from the 9+ to the 13+ charge state, and the average M_r value derived from the five multiply charged ions present in the ESI mass spectrum was 19,266.3 (Fig. 4.1, inset). The excellent mass measurement accuracy, which is usually better than 0.01 % for masses up to 100 kDa [49], makes ESI MS an ideal preliminary method for monitoring the integrity of the therapeutic recombinant protein batches. In case of larger proteins, we observe greater charge states, often in the presence of a dilute acid, due to the presence of more available sites to carry the positive charge (i.e., K-, R-, H-, N-terminus). The simultaneous shift of the observed ion envelope distribution to lower m/z values is also accompanied by a decrease in the spacing between adjacent charge states, thus making the identification of the envelope’s charge-state components difficult. This becomes more complicated in the analysis of a recombinant glycoprotein sample where the inherent complexity of the carbohydrate structure heterogeneity enhances the aforementioned analytical challenge. This complexity is shown in the ESI mass spectrum of the Chinese hamster ovary (CHO)-derived interleukin-4 (IL-4), a glycoprotein containing two N-linked glycosylation sites (Fig. 4.2) [50].

Fig. 4.1

Positive-ion ESI mass spectrum of human recombinant interferon α-2b (INTRON A)

Fig. 4.2

Positive-ion ESI mass spectrum of CHO-derived interleukin-4 (IL-4): Raw spectrum (a) and deconvoluted spectrum (b) showing the individual glycoforms of the glycoprotein. (Reprinted with permission from Wiley [50])

The ESI mass spectrum of CHO IL-4 (Fig. 4.2a) contained three envelopes of multiply charged ions ranging from 8+ to 10+ charge state, with each envelope containing several peaks corresponding to individual glycoforms of the glycoprotein and adducts thereof. This is better depicted in the deconvoluted mass spectrum (Fig. 4.2b), with the mono- and disialylated components (separated by 291 Da) representing the most abundant signals. Other higher M_r components indicated the presence of tri- and tetraantennary glycans containing up to three additional lactosamine units (in-chain mass of 365 Da), whereas satellite signals 98 Da higher were also observed (Fig. 4.2b). These signals probably arise from the attachment of sulfate groups, since sulfate salts were used in the protein isolation process and operating at a higher desolvation potential or using low pH solvents can minimize their formation [51]. Overall, the success of glycoprotein analysis by ESI MS depends on their relative carbohydrate content, with the success decreasing significantly with a relatively high percentage weight of the carbohydrate component. ESI MS analysis of complex glycoproteins by direct infusion often results in broad unresolved signals arising from the large number of different glycoforms and potential salt adducts, along with the ESI multiple charging phenomenon that spreads the signals over a large m/z region. In agreement with that, ESI MS analysis of the 44 kDa ovalbumin containing 4 % carbohydrate was successful [52], whereas glycoproteins with higher carbohydrate content such as the CHO IL-5 (M_r ~ 31 kDa; 15 % carbohydrate) and CHO IL-4 receptor (IL-4R; M_r ~ 38 kDa with 35 % carbohydrate) did not give any ESI signals [53]. Another contribution to the unsuccessful ESI MS analysis is the poor ionization efficiency in the positive ion mode due to the presence of negatively charged glycans, as this is demonstrated in the comparative analysis of recombinant human erythropoietin (rHuEPO) and its asialo counterpart [54]. The use of nano-electrospray ionization (nESI) overcomes this drawback and improves the sensitivity of analysis due to the generation of smaller-sized droplets [55]. Moreover, the interfacing of the nESI source with orthogonal time-of-flight (oTOF) instrumentation [56] has led to better mass measurement accuracy and increased analytical mass range, thus offering new momentum to the ESI MS analysis of glycoproteins. It should be emphasized that the commonly used quadrupole, quadrupole ion trap, and even Orbitrap [57] analyzers have mass range of analysis limited to m/z 2,000 and 4,000 (Orbitrap), which is a significant drawback when larger glycoproteins or non-covalent complexes thereof must be detected; thus, an upper mass limit greater than even m/z 10,000 may be required [58, 59]. This is shown in the nESI mass spectrum of Sf9-derived IL-4R (M_r ~ 30.2 kDa) obtained on an oTOF mass spectrometer, where an extensive series of multiply charged ions up to m/z 3,000 corresponding to two sets of high-mannose glycoforms separated by a fucosylated Man₃(GlcNAc)₂ structure (in-chain mass of 1,039 Da) were observed [51]. Therefore, the improved mass resolving power, sensitivity and extended mass range, has made the oTOF, hybrid quadrupole TOF (QTOF), and recently the ion mobility (IM) [60] TOF as the analyzers of choice for nESI MS analysis of glycoproteins. The use of the IM TOF analyzer is nicely shown in the nESI mass spectrum of the intact therapeutic mAb trastuzumab (Herceptin), which is a humanized monoclonal immunoglobulin γ-1 (IgG1) antibody directed against the HER2/neu receptor, which is over-expressed in about 25 % of all breast cancer patients [61]. In the ESI mass spectrum of trastuzumab obtained on an IM TOF mass spectrometer [62] (Fig. 4.3), an extensive series of multiply charged ions ranging from the 35+ up to the 75+ charge state were observed, and the separation between successive charge states was sufficient to reveal the presence of six glycoform variants. The illustration of these glycoforms is portrayed in the zoomed spectrum for the 53+ charge state (Fig. 4.3b), while their respective assignment is shown in the deconvoluted mass spectrum (Fig. 4.3c). The spectrum clearly reveals the glycoprofile difference between trastuzumab antibodies from different batches (shown in different colors) where the intensity of each glycoform varies.

Fig. 4.3

Positive-ion ESI ion mobility (IM) TOF mass spectrum of the intact therapeutic monoclonal antibody trastuzumab (Herceptin) (a); The 53+ charged ion with the signals corresponding to various glycoforms is annotated (b); The glycoform assignments and the glycosylation heterogeneity of the monoclonal antibody arising from variations in the hexose and fucose content are shown in the deconvoluted mass spectrum (c). (Reprinted with permission from Springer [62])

It should be mentioned that the mass measurement accuracy of the main glycoform is within 1.5–2 Da (~10 ppm) from its theoretical mass value (148,057 Da), an unthinkable achievement prior to the advent of ESI and MALDI MS. The latter is an essential attribute of this method and renders it suitable to distinguish the lot-to-lot heterogeneity in glycosylation profile of the commercially available glycoprotein biopharmaceutical. Glycoprotein heterogeneity can result in an enhancement or loss of the protein’s biological activity, as this has been demonstrated in the case of rHuEPO, where desialylation causes complete loss of its hormonal activity in vivo [63]. In particular, intravenously administered rHuEPO consisting of highly branched sialylated oligosaccharide structures has been shown to result in a plasma half-life of 5–6 h as compared to desialylated rHuEPO, which is cleared within minutes [64].

On the other hand, glycoprotein analysis by MALDI MS yields signals corresponding to protonated molecules (MH⁺) of the individual glycoforms and allows the determination of the heterogeneity for glycoproteins with M_r less than 30 kDa and a relatively low percentage of carbohydrate content. This is clearly shown in the screening of the glycosylation profile for the human soluble urokinase-type plasminogen activator receptor (uPAR) expressed in CHO cells, where the extent and type of glycosylation in its three domains was assessed by MALDI MS [65]. On the contrary, MALDI MS analysis of the Sf9-derived interleukin-5 receptor α-subunit (IL-5Rα) [66] and CHO IL-4R [53] with 17 and 35 % carbohydrate content, respectively, did not provide any information on the type of the glycosylation. In addition, the choice of an appropriate MALDI matrix is very important toward achieving the optimum mass resolving power and separation of the individual glycoform signals [67, 68]. This is shown in the MALDI mass spectra of the Sf9-derived IL-5Rα (in a reflectron and a linear TOF instrument) using different matrices, where the use of the sDHB matrix (2,5-dihydroxybenzoic acid with a 10 % admixture of 2-hydroxy-5-methoxybenzoic acid) in a linear TOF instrument resulted in a more reliable mass measurement due to minimized metastable fragmentation [69] (Fig. 4.4).

Fig. 4.4

Positive-ion MALDI-TOF mass spectrum of Sf9-derived interleukin-5 receptor α (IL-5Rα) obtained with reflectron TOF with sDHB (a) and HPA (b) matrix, and linear TOF instrument using sDHB matrix (c). The asterisk denotes an internal calibrant. (Reprinted with permission from Wiley [69])

Overall, ESI MS analysis of intact glycoproteins has better success over MALDI MS for surveying the individual glycoforms in a glycoprotein biotherapeutic sample and ensuring the homogeneity of the manufacturing batches. Nevertheless, the biggest challenge for the analysis of glycoproteins is their low abundance compared to that of unmodified proteins and the resulting low intensities of the mass spectral signals. This is mainly due to the low ionization efficiency of glycoproteins and the distribution of their signal among the various glycoforms sharing a common peptide sequence, thus rendering their detection an overwhelming task. This can be overcome by performing an enrichment step for the glycoproteins, which eliminates the most abundant unmodified proteins from competing for charge during the ionization process and results in higher ionization efficiencies and increased probability for detecting glycoproteins. The commonly used analytical methods for glycoprotein/glycopeptide enrichment are discussed in Sect. 4.2.2.2. Another promising solution to this problem is coupling of ESI MS with a separation device such as nano-LC [70], CE [71] or CZE [72] that can definitely improve the chances for a successful analysis. This is shown in the analysis of intact rHuEPO and bovine α1-acid glycoproteins by a developed CZE-ESI MS method without any complicated sample treatment, where characterization of the intact glycoforms was provided along with their relative intensities [73, 74]. In addition to the efficient separation of the intact glycoforms, small glycan modifications such as acetylation, oxidation, and sulfation could be successfully characterized. Similarly, high-resolution CE-Fourier transform ion cyclotron resonance (FT ICR) MS analysis was used for the profiling of the intact glycoforms of recombinant human chorionic gonadotrophin (r-RhCG) produced in a murine cell line, which resulted in the identification of over 60 different glycoforms with up to nine sialic acids [75]. These studies suggest that CE-MS can be an important tool for rapid assessment of the recombinant product quality either for product release or for in-process control, and even for demonstrating comparability of a glycoprotein therapeutic biosimilar to the innovator product being replicated.

Moreover, the rapid assessment of glycosylation at the molecular level is invaluable in glycoform screening of glycoproteins involved in certain diseases, such as the human transferrin (Tf) model glycoprotein for congenital disorders of glycosylation (CDG) diagnosis. CE-ESI MS was used successfully for carbohydrate-deficient transferrin (CDT) detection and CDG-type characterization [76]. Comparative analysis of serum samples from healthy and CDG patients by CE-ESI MS (Fig. 4.5) provided partial separation of Tf glycoforms and identification of the carbohydrate-deficient Tf glycoforms in the CDG patients’ serum. It is clearly shown that the Tf glycoforms in the CDG serum correspond to a disialoform containing one free N-glycosylation site (Fig. 4.5e) and another one occupied by a biantennary instead of a triantennary N-linked sialylated glycan (Fig. 4.5f), thus confirming that the sample belongs to a patient who has CDG of type I [76].

Fig. 4.5

Total ion electropherogram obtained for a serum from a healthy individual (a) and a CDG patient (d) under optimized CE-ESI MS conditions. The deconvoluted spectra obtained from the beginning and the end of the Tf peak are shown in (b) and (c), respectively. In case of the CDG patient, the deconvoluted spectra obtained from the two partial resolved glycoforms of Tf provided identification of the carbohydrate-deficient Tf glycoforms in the CDG patients’ serum (e) and (f). The most probable glycan composition is displayed below the deconvoluted mass spectra. (Reprinted with permission from Wiley [76])

4.2.2 Mass Spectrometry and Glycoproteomics

Glycoproteomics involves the study of the glycosylation of proteins, including the structures of the attached oligosaccharide moieties and the identification of the glycosylation sites. There are two distinct classes of protein glycosylation in nature depending on the linkage site. First, the “O-linked” are the ones that are linked to serine (S), threonine (T), or hydroxyproline residues in the protein backbone, and then the “N-linked” which are linked to N residues through an N-acetylglucosamine residue (GlcNAc). Regarding O-glycosylation, a number of monosaccharides attached to S and T have been found, most commonly N-acetylgalactosamine (GalNAc), GlcNAc, xylose, mannose, and fucose [29]. O-glycans are synthesized in a stepwise process that involves single monosaccharide transfer steps, and their biosynthesis takes place after protein N-glycosylation, folding, and oligomerization. O-glycosylation may occur at any S or T residue with no single common core structure or consensus protein sequence. Extended structures from a core GalNAc that are called mucin-type O-glycans are the most frequently occurring [77, 78]. In contrast to O-glycans, N-glycosylation sites can be predicted by the tripeptide sequon Asn-Xaa-Ser/Thr (N-X-S/T, where X is any amino acid except P) [79, 80] (Fig. 4.6). All three types of N-glycans found in mature glycoproteins share a pentasaccharide core (i.e., the trimannosyl core with two N-acetylglucosamine residues (Man₃GlcNAc₂)) because of a common biosynthetic pathway in the endoplasmic reticulum compartment of the cell. This N-glycan Man₃GlcNAc₂ core is common to complex, high-mannose, and hybrid structures as shown in Fig. 4.6.

Fig. 4.6

Classes of N-linked carbohydrate structures sharing a common pentasaccharide core structure, that is, the trimannosyl core with two N-acetylglucosamine residues (Man₃GlcNAc₂). a High-mannose-type; b Complex-type (triantennary); c Hybrid-type. The sugar symbols used throughout this chapter are those adopted by the consortium for functional glycomics (CFG). Circles represent hexoses (Hex) [yellow: Galactose (Gal), green: Mannose], squares represent N-acetylhexosamines (HexNAc) [blue: N-acetylglucosamine (GlcNAc), yellow: N-acetylgalactosamine (GalNAc)], red triangle: fucose, purple diamond: N-acetylneuraminic acid (NeuAc). (Gal: , Man: , GlcNAc: , GalNAc: , Fuc: , NeuAc: )

The high-mannose-type glycoproteins (e.g., ovalbumin) contain two to eight mannose residues added to the pentasaccharide core. Glycoproteins containing complex-type N-structures (e.g., fetuin) exhibit the highest structural variation by having a number of GlcNAc, Gal, Fuc and NeuAc (sialic acid) residues attached to the N-glycan Man₃GlcNAc₂ core, as well as possible extension and/or branching of the outer chains through lactosamine repeats and sialylation. Finally, the hybrid-type glycoproteins combine features from both high-mannose- and complex-type glycans [79]. At this point, it should be emphasized that for both N- and O-glycosylation, there is an inherent microheterogeneity resulting from the array of glycan structures associated with each glycosylation site. Moreover, there is macroheterogeneity due to the fact that not all N-glycan sequons or S/T residues present in the glycoproteins are glycosylated. The end result is a diverse degree of occupancy at different O- or N-linked glycosylation sites with a wide array of structurally different oligosaccharides that generate a complex mixture of glycosylated variants (glycoforms). The variety of these glycoforms depends not only on the polypeptide backbone and the number of putative glycosylation sites but also on the cell type, in which the glycoprotein is expressed, and its development stage. Therefore, characterizing the glycoproteome is a demanding task because of the inherent macro- and microheterogeneity of glycans along with the complex nature of this modification.

4.2.2.1 Top-Down and Bottom-Up Analytical Approaches

Complete structural characterization of a glycoprotein includes the following tasks: (1) characterization of glycans in intact glycoproteins (2) determination of the protein primary sequence and the glycosylation attachment sites (3) characterization of glycopeptides, and (4) structural analysis of chemically or enzymatically released glycans. The M_r determination of intact glycoproteins by either ESI or MALDI MS analysis is successful only for glycoproteins up to 20–30 kDa with a relatively low percentage of carbohydrate content as demonstrated above (Sect. 4.2.1). Even though this accurate M_r measurement is valuable for profiling of intact glycoproteins and providing very useful information on the type and extent of glycosylation, there is no information on the nature and the attachment sites of the glycan chains. Therefore, one needs to cleave the protein into smaller fragments before attempting MS analysis. In the top-down approach, the intact molecule is introduced into the mass spectrometer where limited fragmentation of the ionized protein is induced (i.e., in the gas phase) and the resulting product-ion mass spectra can provide information on the location of the glycosylation sites (or other PTMs) [81, 82]. Even though there are several mass analyzers capable of measuring intact proteins and large ionic fragments (such as TOF, QTOF, FT ICR), the unusually high resolving power (>10⁵) of FT ICR has made possible accurate assignments of ESI charge state and mass, even for MS/MS of intact proteins [83, 84]. Such top-down methods have proven especially powerful in stability and formulation studies of intact antibodies with M_r ~ 150 kDa used as therapeutics in the biopharmaceutical industry. Nevertheless, FT ICR instruments have not become standard analytical tools for the characterization of recombinant biopharmaceuticals mainly due to the high cost of acquisition and maintenance. For that reason, the most commonly followed MS-based approach for characterization of a recombinant biopharmaceutical involves the enzymatic digestion of the glycoprotein (usually with trypsin or another endoprotease) followed by the separation/analysis of the resulting peptide digests by LC–MS/MS [41, 85, 86] or CE-MS/MS [87] and MALDI MS [88] (bottom-up approach). In case of purified proteins or simple mixtures thereof, LC–MS or MALDI MS analysis of the proteolytic mixture provides M_r information on the peptide components. The advantage of MALDI-TOF MS is the simplicity of the spectra, which contain usually intense protonated (MH⁺) or sodiated signals corresponding to the enzyme-generated peptides. Further, protein structural information can be deduced by carrying out LC–MS/MS analysis of the enzyme-generated peptides. Peptide identification is performed through a direct search of the M_r measured values and the tandem MS-derived fragment ions (sequence tags) [89] against a protein sequence database (peptide fingerprinting). The general experimental workflow comprising the commonly employed approaches in glycoproteomic analysis is shown in Fig. 4.7. Of course, the nature of the glycoprotein sample determines the number of the necessary steps needed in order to determine site-specific glycosylation and heterogeneity.

Fig. 4.7

The general experimental workflow comprising the commonly employed approaches in glycoproteomic analysis. The top-down approach starts with the analysis of intact glycoproteins, while the bottom-up analytical approach begins with proteolytic or chemical cleavage of the glycoprotein, followed by mapping of the generated glycopeptides by an assortment of LC–MS and tandem MS techniques

In general, MS mapping of the enzyme-generated peptide mixtures provides not only confirmation of the expected protein sequence but also identification of any existing modifications, including the glycosylation attachment sites. In addition, unexpected mass spectral signals can provide insights into the glycosylation profile of the protein, taking into consideration the known N-glycan structures (Fig. 4.6). Nevertheless, there are several problems associated with the bottom-up approach. The major problem arises from the fact that many glycoproteins are resistant to enzymatic proteolysis (e.g., trypsin or S. aureus V8 protease) due to the presence of the attached glycans near the proteolytic site, thus requiring an additional specific enzymatic proteolysis. In addition, the resulting mixture of peptides and glycopeptides could complicate the analysis because glycosylation strongly diminishes the ionization efficiency of the peptide [90, 91], especially when the glycans are terminated with the negatively charged sialic acid moiety [47]. This problem becomes more significant considering that the glycopeptides are in much lower abundance than the peptides from the same glycoprotein, and the glycopeptide signals are distributed over several peaks due to the glycan heterogeneity and multiple adduct ion formation. However, several enrichment methods (either in parallel or sequentially) prior to glycoprotein analysis can be used to compensate for the low abundance of glycopeptides (and glycoproteins) and the presence of multiple glycan structures (heterogeneity) [92]. The use of glycoprotein enrichment methods can bypass the aforementioned obstacles in glycoprotein analysis by achieving exclusion or reduction of the most abundant unmodified peptides from the analysis, thus improving the ionization efficiency of the low-abundance glycopeptides, which do not have to compete for charge during the ionization process with unmodified peptides.

4.2.2.2 Glycopeptides Enrichment Methods

Enrichment of glycoproteins and glycopeptides can be achieved by using the natural affinity of lectins for their glycan “handles” [93], whereas other analytical methodologies based on general physical and chemical properties of glycopeptides have been employed, such as size-exclusion chromatography (SEC) [94], hydrophilic interaction chromatography (HILIC) [95–97] or graphitized carbon columns (GCC) [98, 99]. A rough classification of the commonly used enrichment techniques in glycopeptides analysis can be made into chemical [100, 101] and chromatographic methods (such as affinity chromatography [102–104], LAC [93], immunoaffinity chromatography [105], SEC [94], hydrophilic phases [96, 97] and GCC [99]).

4.2.2.2.1 Lectin Affinity Chromatography

Lectins are proteins originating from plants, fungi, bacteria, or animals that express a special affinity toward glycans [106] and thus are used for glycopeptide/glycoprotein isolation from complex mixtures after being immobilized onto various solid supports such as silica [107], agarose [46], resins, magnetic beads, and affinity membranes. These are used in different arrangements, such as columns [108–110], tubes [46], and microfluidic chips [111]. Lectins generally interact with specific motifs in a glycan and demonstrate selectivity for different oligosaccharides and broad range of specificity [112], thus enabling glycoprotein/glycopeptide isolation from a complex protein mixture along with glycoform pre-fractionation. Widely used lectins include concanavalin A (ConA) [113, 114], which binds glycan residues containing mannose and glucose and affords broad selectivity (i.e., high-mannose, hybrid, complex biantennary [115]), wheat germ agglutinin (WGA), which presents selectivity for GlcNAc and NeuAc, and Jacalin (JCA), which expresses affinity against galactose (b1-3) GalNAc and O-linked glycoproteins.

Various analytical strategies have been proposed for the isolation and pre-concentration of glycoproteins/glycopeptides prior to MS analysis [93]. In summary, the sample enrichment using lectin columns can be performed before or after the protein mixture digestion by loading the sample onto the columns under high-ionic-strength buffers to prevent non-specific retention. The same loading buffer containing a displacer (haptene saccharide) is used to elute the captured glycopeptides/glycoproteins, which can then be subjected to MS analysis.

There are two principal enrichment methodologies based on LAC: Serial Affinity Chromatography (SLAC) [116] and Multi-Lectin approach (M-LAC) [117]. The first one uses a serial set of lectin columns with different specificity, thus enabling the sequential selective binding of various glycan moieties of a peptide or protein mixture. SLAC has proven to be a powerful tool for rapid and primary elucidation of glycans’ structural features, especially when columns with broad (ConA, WGA, or JCA) and narrow selectivities (also known as “structure-specific affinity selectors,” i.e., Sambucus nigra agglutinin, SNA) are combined [118]. The SLAC approach was used in the characterization of a prostate-specific antigen in human prostate cancer [119]. Furthermore, coupling LAC with advances in stable isotopic labeling has been successfully applied for the comparative analysis of sialylated proteins [120], thus providing a valuable tool for exploring the glycosylation sites of the whole proteome as well as an excellent tool for biomarker discovery. On the other hand, the M-LAC approach uses a single column (multi-lectin column) containing various lectins with broad specificity (i.e., ConA, WGA, JCA), thus enabling the comprehensive isolation of glycoproteins/glycopeptides from a complex mixture covering an extended dynamic range. This approach was used in the study of glycoproteins in human serum [117, 121] and plasma [122].

Integrated analytical platforms combining LAC with various separation techniques have been developed lately in order to overcome the low-throughput drawback of the off-line procedures. Such methodologies include a microfluidic chip [111] containing a polymeric monolithic column with immobilized pisum sativum agglutinin (PSA), an integrated glass microchip [123] for online tryptic digestion of glycoproteins in the first channel, followed by selective enrichment of resulting peptides through ConA in the second channel. The eluted fractions were subjected to CE and nano-LC–MS analysis employing capillary polymethacrylate monolithic columns with immobilized ConA and WGA [109] allowing large volume injection and adequate sensitivity. Similarly, a fully automated LAC system coupled online to ESI MS with silica-based lectin microcolumns [108] demonstrated high-binding capacity and excellent reproducibility, whereas a variation of this platform with SLAC [124] was proved to be superior over the M-LAC approach for the selective enrichment of small volumes of blood serum [115].

4.2.2.2.2 Immunoaffinity Chromatography

Immunoaffinity (IA) enrichment protocols for glycoproteins/glycopeptides rely on the unique specificity of the antibody–antigen interaction and enable the highly selective adsorption of a target analyte through the covalent attachment on a properly functionalized solid support containing an affinity ligand [125]. This can be performed either by the covalent attachment of antibody fragments via proper chemistries that provide correct orientation of the fragment, or by immobilization of a secondary binder molecule. The elution of the bound ligands is achieved by lowering the pH of the eluting buffer to pH 1–3, by using chaotropic salts, or by using polarity-reducing agents in order to weaken the antibody–antigen hydrophobic interactions. Although the IA enrichment approach has been mainly used for off-line targeted glycoproteomics [105], an online integration of this technique with CE was employed for the pre-concentration of rHuEPO in diluted solutions [126]. This integrated platform has demonstrated high loading sample capacity and good separation efficiency of the glycoforms.

4.2.2.2.3 Porous Graphitized Carbon Chromatography

Porous graphitized carbon chromatography (PGC) has been employed for the separation of oligosaccharides, in their native form as well as after derivatization, based on a retention mechanism driven mainly by hydrophobic and electrostatic stacking interactions. The oligosaccharide analytes are eluted in order of increased size, and structural isomer resolution is often provided [47]. In addition to separation, PGC has been used for the selective enrichment of glycans and glycopeptides. An off-line approach combining solid-phase extraction (SPE) with PGC cartridges was used to concentrate and pre-fractionate pronase glycopeptides and glycans prior to MALDI-TOF MS analysis [127]. An automated variation of the aforementioned approach for glycoprotein analysis has been reported combining digestion, extraction, and separation processes in one analysis [128]. This integrated platform employs a pronase-based chromatographic bioreactor for the in situ rapid digestion of glycoproteins, an online SPE of the produced glycopeptides with a PGC trap column, and separation by LC–MS/MS. This system allowed the direct sequencing of the glycans and peptides along with simultaneous characterization of the glycan composition and localization of the glycosylation site.

4.2.2.2.4 Chemical Derivatization Methods

In addition to the affinity techniques described above, that do not change the structure of the modification and the peptides/proteins, several chemical methods specific to the glycan moieties have been used for the detection and the purification of glycosylated proteins. Most of the chemical derivatization strategies use two basic reactions: (1) the Schiff base reaction of aldehydes with a hydrazine [129–131], and (2) a Staudinger ligation between a phosphine and an azide [132, 133]. However, most of these derivatization methods provide peptide/protein identification without much information about the site or the structure of glycosylation, mainly due to inadequate search algorithms and the occasional modification of the glycan structure [112].

One of the strategies using the Schiff base reaction is the O-GlcNAc ketone enrichment method [134], where a chemo-enzymatic approach using an engineered β-1,4-galactosyl transferase is employed to transfer a ketone containing substrate onto O-GlcNAc-modified proteins. A Schiff base reaction was used to biotinylate the ketones with biotin-hydrazine and subsequently the O-GlcNAcylated peptides/proteins were captured on a streptavidin affinity column. This methodology was successfully used for the identification of the cAMP-responsive Element-Binding Factor (CREB), a low-abundance protein with two known O-GlcNAc sites, in a whole cell lysate [135]. Another derivatization enrichment approach for the glycoproteome and especially for N-glycosylation, is the Periodate-acid-Schiff (PAS) reaction using an iminobiotin hydrazide via the Schiff base reaction [129]. The derivatized peptides/proteins are affinity purified on a streptavidin column and analyzed by MS. This reaction exploits the unique vicinal diol functionality of glycans, thus oxidizing these diols to aldehydes without affecting any other amino acid apart from M, which is oxidized to its sulfoxide analog. This approach provides important information regarding N-glycosylation site modifications and has been used for high-throughput quantitative analyses [136]. In addition, it is an extremely versatile process for proteins and peptides, as different coupling agents such as biotin hydrazides and digoxigenin hydrazides can be incorporated. However, the major disadvantage of the PAS strategy is the heterogeneous modification of the glycan structures by an undefined number of hydrazide tags, thus necessitating PNGase cleavage of the glycans in order to sequence the peptide backbone. In this way, all information pertaining to the glycan structure is lost and only N-glycosylation sites can be determined.

In a modification [133] to the standard Staudinger reaction (a reaction of an azide with a phosphine), the intermediate aza-ylide formed in the standard Staudinger reaction reacts with an electrophilic trap to form an amide bond with a compound that is biotin tagged. This reaction is biologically unique as neither phosphines nor azides occur in biomolecules and also offers the possibility to design phosphines in order to incorporate a wide variety of tags, such as fluorescent probes and affinity tags [137, 138]. A tagging-via-substrate (TAS) strategy based on a tag attached to the modification substrate was used for the identification of O-GlcNAc glycosylated proteins [117, 124], as well as for the detection and isolation of other PTMs in proteins, such as farnesylation [139]. Another derivatization method that has been used for the enrichment of O-linked β-GlcNAc is β-elimination followed by Michael addition with dithiothreitol (BEMAD) [140]. BEMAD has also been used to quantitate both O-glycosylated and O-phosphorylated peptides [141].

4.2.3 Determination of Site-Specific Glycosylation and Heterogeneity

The complete characterization of a glycoprotein biopharmaceutical involves the analysis of the glycan structures that are expressed on the glycoprotein of a given organism or cell line, the identification of the proteins that express these glycans, as well as the individual glycosylation sites on each protein [39, 41]. MS and tandem MS analysis of glycopeptides usually after chromatographic or electrophoretic separation, either online or off-line, holds a central role in all the strategies for glycoproteomic analysis [142] (Fig. 4.7). The most commonly followed experimental approach for providing a detailed glycoprotein mapping involves the analysis of enzymatically derived glycopeptides by fast atom bombardment (FAB) [22, 143], MALDI [144] or LC-ESI MS and MS/MS [50, 85]. This MS mapping identifies most of the expected peptide signals (peptide fingerprinting), whereas any new, unexpected mass spectral signals may correspond to glycopeptides. In a similar off-line strategy, the isolated fractions are mapped by ESI or MALDI MS and MS/MS approaches. Nevertheless, there are several issues related to these MS approaches, such as the potential deglycosylation of glycopeptides in the gas phase combined with the low ionization efficiency and low abundance of the glycopeptides compared with the peptides derived from the proteolytic digestion. One of the remedies to ensure the appearance of glycoproteomic information within the copious proteomic data is enrichment of glycoproteins and/or glycopeptides prior to analysis (as discussed above). Another way to overcome this difficulty is the carbohydrate removal from the glycoprotein by base-catalyzed β-elimination for O-linked glycans or digestion with PNGase F (N-Glycanase) for N-linked glycans. The former leads to the conversion of S and T residues to A and α-aminobutyric acid sequences, respectively (i.e., loss of 16 Da), whereas the latter converts the glycosylated N residues to D (i.e., increase of 1 Da). In the MS mapping of the enzyme-generated peptide mixture of the deglycosylated protein, the former O- and N-glycosylated peptides can be readily identified by the appearance of new mass spectral signals at lower m/z (for O-linked sugars) or higher m/z (for N-linked sugars) than those of the unglycosylated peptides [145]. This mass difference can be magnified by carrying out the N-Glycanase reaction in fully or partially (50 %) ¹⁸O-labeled glycosylated N residues, which results in characteristic doublets separated by 2 Da [146]. These doublets can be used to locate the modification site and to determine the degree of occupancy at each N-linked glycosylation site. Another approach for N-linked glycans involves the release of the high-mannose- and hybrid-type oligosaccharides by digestion of the glycoprotein with endoglycosidase H, leaving a GlcNAc residue attached to the peptide’s N residue. That results in the detection of peptides having M_r values 203 Da higher than that of the respective unglycosylated peptides. Glycosylation sites containing complex-type glycans are unaffected by the endoglycosidase H treatment. This approach was employed in the FAB carbohydrate mapping of the major envelope glycoprotein gp120 of HIV type 1 [147] and recombinant tissue plasminogen activator (rtPA) [148]. On a similar approach, glycoprotein mapping of CHO rHuEPO was facilitated by removal of terminal NeuAc residues with neuraminidase followed by LC-ESI MS analysis of the enzyme-generated peptide fragments of asialo CHO rHuEPO [54]. rHuEPO contains three N-glycosylation sites at N-24, N-38, and N-83 and a single O-glycosylation site at S-126; the glycans account for up to 40 % of the total molecular mass. This LC–MS mapping provided information on the microheterogeneity of the carbohydrate structures, which is associated with the presence or absence of lactosamine extensions and varying levels of O-acetylated NeuAc residues. Similarly, comparative LC-ESI MS tryptic mapping of untreated and neuraminidase-treated rtPA allowed the identification of the attachment site of two hybrid-type carbohydrates on one of the tryptic peptides [149]. The same analytical protocol was applied in the characterization of a rtPA mutant with an additional glycosylation site (T103N), where two new complex-type carbohydrate chains have been observed [149]. An analogous LC-ESI MS/MS approach combined with a multi-enzymatic digestion strategy was employed for the characterization of the glycosylation occupancy in the generic variant of rtPA (TNK-tPA), which was approved for treatments of acute myocardial infarction and ischemic stroke [150]. TNK-tPA has the same amino acid sequence as natural human tPA except for the three substitutions: T103N, N117Q, and AAAA for KHRR (296–299) which lead to longer half-life and higher fibrin activity than those of tPA. Nevertheless, differences in the glycosylation occupancy at N184 along with different extents of deamidation at N184 and oxidation at M207 have been observed between the therapeutic biosimilar and the innovator product, thus raising concerns as to its bioequivalence.

In the case of CHO IL-4, comparative LC–MS tryptic and V8 protease mapping of CHO IL-4 and its N-Glycanase-treated protein revealed that the N residue in the sequon N³⁸TT was glycosylated rather than the other potential site at N¹⁰⁵QS. We should point out that the presence of carbohydrate often provides shielding of a neighboring proteolytic site, thus leading to the incorporation of the adjacent peptide fragment, as demonstrated by the incorporation of the T5 tryptic glycopeptide into the adjacent disulfide-linked peptide T4–T10 of CHO IL-4 [50]. When ESI MS/MS approaches are incorporated in the analysis of the LC- or CE-separated enzymatic fragments of a glycoprotein, the identification of glycopeptide-containing chromatographic fractions is facilitated by the appearance of several diagnostic fragment ions. CID product-ion spectra of ESI generated glycopeptides in a variety of instruments, such as triple quadrupole, ion trap (IT), and QTOF, are dominated by fragmentation of glycosidic linkages thereby revealing predominantly information on the composition and sequence of the glycan moiety. Glycopeptide marker ions under CID conditions are low-mass sugar-specific oxonium ions (B-type fragmentation in the Domon and Costello nomenclature [151]) of m/z 162 for Hex⁺, m/z 204 for HexNAc⁺, m/z 274 and 292 for NeuAc⁺, m/z 366 for Hex-HexNAc⁺, and m/z 657 for NeuAc-Hex-HexNAc⁺. Scanning for these diagnostic fragment ions in the “precursor ion” mode on triple-quadrupole mass spectrometers can selectively identify the glycopeptides within the enzymatic digest mixture, whereas screening of constant neutral losses of terminal monosaccharides could also pinpoint the glycopeptides. Selected ion monitoring (SIM) experiments can also be performed for glycopeptide identification with IT and QTOF mass analyzers. In cases, where MS/MS is not available, these low-mass glycopeptides marker ions can be generated by either “in-source” fragmentation of ESI-produced ions [50, 149, 152] or post-source decay (PSD) of MALDI-produced ions [153]. In the former, increasing the source entrance potential into the mass spectrometer, which controls the collision excitation and the extent of fragmentation, induces the fragmentation. This online LC–MS “in-source” CID mapping of glycopeptides utilizes both low and high source potentials and monitoring of the resulting sugar-specific oxonium ions. In case of complex/hybrid or high-mannose structures, monitoring of the oxonium ions at m/z 204, 274/292, 366 and 657 has allowed the fast glycan profiling in the LC-ESI MS analysis of the trypsin-treated CHO rTPA [154] and CHO IL-4 [50] without having to search each individual mass spectrum for glycopeptide-characteristic patterns. In the case of rTPA, this method allowed the identification of a low-level novel N-glycosylation at N142, which is part of an atypical N-Y-C consensus motif. Although this site is only 1 % occupied by predominantly biantennary hybrid structures, it was readily detected by this sensitive LC-ESI MS tryptic mapping approach. In the case of CHO IL-4, the observation of the glycopeptides marker ions at m/z 274, 366 and 657 revealed the presence of sialylated complex-type N-glycans in the specific chromatographic fraction. In addition, the mass separation of the signals within the triply and quadruply multiply charged ion envelopes revealed the presence of mono- and di-sialylated glycoforms (291 Da apart) along with higher M_r components containing additional lactosamine units (365 Da apart) owing to the presence of extended arms or branching. Similarly, this rapid glycopeptide screening approach was applied to other mammalian-cell-derived proteins, such as the Sf9-derived IL-5Rα, where this low/high “in-source” fragmentation allowed the identification of all glycopeptide-containing fractions in the LC-ESI MS tryptic peptide map of Sf9 IL-5Rα (Fig. 4.8). This method allowed the identification of four glycosylation sites in Sf9 IL-5Rα out of the six potential sites fulfilling the N-glycosylation consensus sequence [66].

Fig. 4.8

Total ion current (TIC) chromatogram obtained from the reversed-phase LC-ESI MS analysis of the Sf9 IL-5Rα tryptic digest using “in-source” fragmentation. The tryptic fractions containing glycopeptides are labeled, as inferred by the detection of the low-mass diagnostic ions at m/z 162, 204, 274, 366, and 657

The ESI mass spectrum of one glycopeptide-containing fraction (Fig. 4.8, peak 10) showed signals corresponding to doubly and triply charged tryptic glycopeptides containing a Man₉(GlcNAc)₂ high-mannose carbohydrate (Fig. 4.9). All these glycopeptides contain the N196 glycosylation site and the M_r values of the respective glycoforms differ by 162 Da due to an extensive heterogeneity in the Man () content, as shown in the deconvoluted mass spectrum (Fig. 4.9, inset).

Fig. 4.9

Positive-ion ESI mass spectrum of Sf9 IL-5Rα tryptic glycopeptide component (Fig. 4.8, TIC peak 10). The deconvoluted mass spectrum (shown in the inset) indicates the presence of a high-mannose carbohydrate component with M_r value of 3366. The asterisk and cross-denoted peaks in the doubly and triply charged ESI ion envelopes correspond to two glycoforms with M_r values 3204 and 3042, respectively, arising from glycoform heterogeneity due to variations in the mannose () content

The assignment of the putative glycan structures to the experimental masses with a high degree of confidence is made possible by the excellent mass measurement accuracy provided by ESI MS analysis. Corroborative information on the composition and sequence of the attached glycans can be attained from MS/MS analysis of the glycopeptides, because CID tandem mass spectra of glycopeptides contain mainly fragments arising from glycosidic bond cleavage [155]. In the analysis of the therapeutic glycoprotein BRP 3 EPO by a combined anion-exchange chromatography (AEC)—ultra-performance liquid chromatography (UPLC) MS/MS approach, tetra-antennary glycans with up to four NeuAc and up to five poly-N-acetyl lactosamine extensions were observed at the glycosylation sites N24 and N83, whereas biantennary glycans were the major structures at N38 [156]. The presence of these large repeating glycan motifs although at low levels may infer additional functional interactions for EPO and may be beneficial in terms of immunogenicity. A more detailed characterization of N-glycopeptides, especially in terms of the peptide sequence, can be obtained by an alternative approach combining MS/MS and MS³ experiments in an IT MS [142]. The glycopeptide ion is selected and fragmented, and the peptide ion carrying a single GlcNAc (which is often the most abundant ion) is subjected to a second fragmentation cycle resulting in extended fragmentation of the peptide moiety into b- and y-series ions, thus allowing the deduction of the glycan attachment site. MS/MS analysis of N-glycopeptides with QTOF mass analyzers at low collision energy exhibited mostly cleavages of glycosidic linkages providing information on the glycan moiety [157]. Nevertheless, CID mass spectra at elevated collision energies resulted in a significant level of b-type and y-type peptide fragmentation, thus allowing identification of the glycosylation site. The potential of the nESI QTOF MS/MS in the characterization of O-glycopeptides has also been demonstrated in the analysis of mucin-type glycopeptides with S- or T-linked O-glycans [88, 158] where information on the structure and the attachment site of the O-glycan has been provided based on the b-type and y-type peptide ions comprising the glycan attachment site.

Alternatively, the development of the complementary mass spectrometric fragmentation techniques of electron-capture dissociation (ECD) [24, 159] and electron-transfer dissociation (ETD) [25] has expanded the analytical options for mapping the modification sites of both N-glycosylation and O-glycosylation. In the ECD technique, which is mainly restricted to FT ICR analyzers, multiply protonated peptide ions are irradiated with low-energy electrons (<0.2 eV) and undergo fragmentation. On the other hand, ETD can be combined with IT, QIT, and Orbitrap analyzers and peptide fragmentation is generated through gas-phase electron-transfer reactions from singly charged anions (e.g., anions of fluoranthene, sulfur dioxide) to a multiply charged peptide/glycopeptide. Unlike the traditional MS/MS techniques, both ECD and ETD appear to retain labile PTMs and induce fragmentation of the peptide backbone with minimal loss of the glycan moiety. ECD and ETD of glycopeptides result in the cleavage of the amine backbone (N–C_α) to generate preferentially c′ and z• fragments ions (nomenclature of Zubarev and co-workers [160]). The intact oligosaccharide moieties are retained in the fragment ions containing the site of glycosylation. Consequently, ECD and ETD represent excellent tools for the localization of modification sites in post-translationally modified proteins [161–163], and there have been few reports of using theses techniques in the characterization of N-linked [142, 164] and O-linked glycopeptides [162, 165]. This is nicely shown in the ESI tandem MS analysis of a tryptic glycopeptide (S295-R313) from horseradish peroxidase (HRP) containing a core-fucosylated and core-xylosylated trimannosyl N-glycan attached to the N298 residue (Fig. 4.10) [142]. The [M+3H]³⁺ ion at m/z 1119 was subjected to CID fragmentation which led to preferential cleavage of glycosidic linkages rather than polypeptide bonds (Fig. 4.10a), thus providing information primarily on the composition and sequence of the glycan moiety. On the contrary, ETD ion activation of the [M+3H]³⁺ ion yielded the cleavage of the peptide backbone with no loss of the glycan moiety, thus leaving the N-glycan modification on the N298 residue intact and providing complete peptide backbone sequence through the observed c′- and z• -ion series (Fig. 4.10b).

Fig. 4.10

Positive-ion nESI tandem mass spectrum of the RP-HPLC-purified tryptic glycopeptide Ser295-Arg313 from horseradish peroxidase (HRP) containing a core-fucosylated and core-xylosylated trimannosyl N-glycan attached to N298 residue. The [M+3H]³⁺ ion at m/z 1119 was subjected to CID (a) and ETD (b) fragmentation which led to cleavage of the glycosidic bonds and the peptide backbone, respectively. (Reprinted with permission from Elsevier [142])

Therefore, the use of both CID and ETD ion activation in the LC–MS analysis of glycopeptides has allowed the characterization of both glycan structure (CID-MS/MS) and peptide sequence/site attachment (ETD-MS/MS) within the same LC–MS run. Similarly, use of LC–MS and the ETD and CID fragmentation techniques allowed the identification of two distinct O-glycopeptide structures and three glycosylation sites from the secreted amyloid precursor protein (sAPP695) expressed in CHO cells [166]. This de novo characterization of unknown O-glycosylation sites was extremely challenging due to the large number of S and T residues (27 S and 39 T residues) contained in the protein sequence of the APP fragment. In a modified strategy, LC–MS combined with CID, ETD, and CID of an isolated charge-reduced species derived from ETD was employed to determine the peptide backbone sequence and the site of modification for an O-linked glycosylated peptide fragment of rtPA at the low femtomol level [167].

In case of glycoprotein mapping by MALDI MS, the intense protonated (MH⁺) glycopeptides signals are much more stable in CID than the multiply protonated glycopeptide species obtained by ESI. Although PSD, as well as CID, is used for MS/MS of glycopeptides, precise analysis of fragment ion peaks often seems to be difficult because of preferential and fast deglycosylation, and the limited peptide sequence information [168]. Therefore, fragmentation of these glycopeptide ions by metastable dissociation in a MALDI-TOF/TOF MS or by CID in a MALDI QTOF instrument is performed at higher energies. MALDI-TOF/TOF MS of N-glycopeptides results in a set of cleavages at or near the innermost GlcNAc residue, with the peptide moiety retained in all the fragment ions. In addition, peptide bond cleavages next to the fragmentation of glycosidic bonds are observed (predominantly b-type and y-type ions), which provide useful peptide sequence tags [169, 170]. All these fragments comprising the N-glycosylation site retain the attached glycan, thus confirming the glycan attachment site. Similarly, MALDI-TOF/TOF MS of O-glycopeptides generate fragmentation patterns from the glycopeptides precursor ions (b- and y-series ions), which can be used for identification of O-glycosylation sites as it was demonstrated in the case of mucin-type glycopeptide derivatives [171].

At this point, we should point out that parallel glycomic analyses for providing information on the linkage, branching points, and configuration of the constituent monosaccharides (microheterogeneity) are also essential in the whole glycoproteomic strategy. In general, the glycans are released by enzymatic or chemical digestion of the glycoprotein or the glycopeptide mixture, undergo permethylation and then subjected to a range of techniques, selected upon the level of analysis to be carried out, that is, fingerprinting, linear sequencing, linkage, branching, or quantitation of monosaccharides [172, 173]. In one of the followed approaches, the permethylated glycans are subjected to LC–MS analysis and the supplied mass spectral information on the specific glycans and their relative amounts can be compared and matched with data at the glycopeptide and overall glycoprotein levels (Fig. 4.7). Incorporation of MALDI-TOF and ESI tandem MS can definitely enhance the analytical potential for tackling complex glycobiology structural issues [43]. Further information on the carbohydrate secondary structures can be provided by well-established methods in structural glycobiology such as X-ray crystallography and especially 2D nuclear magnetic resonance (NMR) analysis [174, 175], albeit the requirement for highly purified glycans and large amounts of sample.

4.2.4 Bioinformatics Tools for Glycoprotein Analysis

Because of the extreme glycan heterogeneity, interpretation of the data produced from the aforementioned glycoproteomic approaches and glycopeptide identification through a comprehensive large-scale data analysis is a challenging task. The development and use of informatics tools and databases for glycobiology research has increased considerably in recent years [176], even though the progress of these tools for glycobiology and glycomics is still in its infancy compared to those already used in genomics and proteomics. Even though, the automated identification of proteins from MS and MS/MS spectra is now almost routine by using informatics tools such as Mascot (http://www.matrixscience.com/), there is lack of rapid and accurate automated tools for retrieving structural information from MS data in case of glycoproteomics. The MS and MS/MS-derived information should be searched for putative glycopeptides predicted by comparison with other glycoconjugate structures derived through the same biosynthetic machinery in other closely related organism, cell line, or tissue. Nevertheless, the inherent complexity of the glycan structures combined with the wide range of techniques employed in their study renders the development of similar automated computational tools a formidable task [177]. In addition, the lack of libraries of glycan sequences similar to the SWISS-PROT protein databank makes matters more challenging. It should be emphasized that more than half of all proteins are glycosylated, based on the analysis of well-characterized proteins deposited in the SWISS-PROT databank [28].

In case of proteomics, bioinformatics tools essentially utilize sequences of the building blocks of proteins (20 amino acids), which are always linked in a predicted linear way in order to provide automated protein identification from MS and tandem MS data. On the contrary, carbohydrates are structurally diverse as their building blocks, the monosaccharides, may be connected in various ways to form branched structures, thus complicating their digital encoding. Moreover, in contrast to protein expression, glycosylation is a non-template-driven synthetic process where multiple enzymes are involved and the final glycoprotein product depends on the type of enzymes expressed in the cell that synthesizes the glycoprotein. The development of bioinformatics methods has mainly found applications in glycosylation analysis, glycomics, glycan structure analysis, glycan biomarker prediction, and glycan structure mining (e.g., using lectins that recognize a certain glycan [178]). In the glycosylation analysis and the prediction of glycosylation binding sites on proteins, the first step is the selective search of protein databases for proteins containing only the consensus sequence for N-linked glycosylation. Several software platforms have been developed for the identification of intact N-linked glycopeptides, such as GlycoMod [179], GlycoPep DB [180], Cartoonist [181], Peptoonist [182], and Glyco-Miner [183]. These methods can be used mainly for glycopeptides generated from specific enzymes, for example, trypsin or endoproteinase Glu-C, whereas GlycoX [184] can be used for interpretation of mass spectra obtained from non-specific proteases, such as protease K. Cartoonist is one of the earlier developed glycomic MS interpretation approaches containing a library of several hundred archetype glycans derived from information about biosynthetic pathways and employing a set of rules to modify these structures. Cartoonist incorporates the same assumptions used by human expert in the annotation of MS data, and it is used to automatically annotate N-glycans in MALDI mass spectra with diagrams or cartoons of the most possible glycans consistent with the observed mass values. Peptoonist [182] uses MS/MS data to identify glycosylated peptides in LC-ESI MS runs of enzymatically digested glycoproteins and MS data to identify the N-glycans present on each of those peptides. On the other hand, the GlyDB [185] approach has been developed to address the need for structure annotation of N-linked glycopeptides in the LC-ESI MS analysis of glycoprotein proteolytic digests. The annotation of low-resolution tandem MS spectra of N-linked glycopeptides arising from low-energy CID, where cleavage along the glycosidic bonds occurs preferentially, is based on matching experimental spectra to theoretical spectra generated by a linearized database of glycan structures using the established search engine SEQUEST. Similarly, GlycoPep ID [186] is a web-based tool used to identify the peptide moiety of either sialylated, sulfated, or both sialylated and sulfated glycopeptides, by correlating the product ions of suspected glycopeptides to a peptide composition. Following the identification of the peptide portion, the mass of the remaining segment can be attributed to the carbohydrate component.

Even though the development and use of informatics tools and databases for glycobiology and glycomics research has increased significantly in recent years, it has lagged behind the development of similar tools for genomics and proteomics. This drawback arises from the lack of comprehensive and well-organized compilations of glycan sequences and efficient automatic assignment procedures for high-throughput analysis of glycans. Most of the aforementioned library-based sequencing and N-glycopeptide identification tools for MS data interpretation are not publicly available; they have their own standards, databases and/or run on a special hardware platform. Moreover, the independently developed database with their own format and language along with the absence of publicly available databases with carefully assigned MS spectra of glycans hinders the development of efficient scoring algorithms. Therefore, rules should be established for the standardization of the structural description of glycans and the deposit of glycan structures and the associated glyco-related data in databases of complex glycan structures. In addition, the deposit of complex glycan structures and glyco-related data in generally accepted databases should be maintained by well-recognized international institutions such as NCBI (www.ncbi.nlm.nih.gov) and European Bioinformatics Institute (EMBL-EBI, www.ebi.ac.uk), which house genome sequencing data (GenBank) and protein related databases, respectively. It is also essential to ensure the intercompatibility of the related data formats, in order to facilitate data exchange between different databases and efficient cross-linking and referencing thereof between various projects.

Toward this direction, the EU FP6-funded EUROCarbDB project (http://www.ebi.ac.uk/eurocarb/home.action) was an initiative to create the technical framework where interested research groups could feed in their complex glycan structural data, which would be archived and maintained at the EMBL-EBI. Other most prominent publically available glycan-related databases are the Consortium for Functional Glycomics (CFG) relational database (http://www.functionalglycomics.org/glycomics/common/jsp/firstpage.jsp), the Kyoto Encyclopedia of Genes and Genomes glycome informatics resource (KEGG GLYCAN) (http://www.genome.jp/kegg/glycan/), and Glycosciences.de (http://www.dkfz.de/spec/glycosciences.de/sweetdb/index.php). Finally, genomic/proteomic findings need to be integrated with biomedical studies where glycan structures can serve as biomarkers for specific diseases or malfunctions [187], like the ones provided by the KEGG resources [188–190].

4.3 Disulfide Bond Formation

4.3.1 MS Determination of Disulfide Bonds

Even though glycosylation enjoys more popularity in the PTM literature, disulfide bond formation is one of the most common PTMs playing a critical role in establishing and stabilizing the three-dimensional structure of proteins [6, 191]. The physiological and pathological relevance of disulfide bonds to diseases has been recognized in several cases, such as tumor immunity [192], neurodegenerative diseases [193], and G-protein receptors [194]. These cross-linkages between the sulfhydryl groups of two C residues can be either intramolecular or intermolecular. The former stabilize the tertiary structures of proteins, while the latter are involved in stabilizing quaternary structures of proteins [195, 196]. For protein therapeutics, the generation of correctly folded recombinant proteins is of paramount importance. Difficulties in folding recombinant protein products are common from E.coli cell line, thus resulting in loss of specific activity compared to the native material. Similarly, over-expression of proteins in CHO cell line leads to disulfide scrambling. Therefore, there are significant efforts to develop reliable methods for mapping disulfide bonds in therapeutic proteins, thus ensuring drug quality. The determination of disulfide bond arrangements of proteins not only provides insights into protein activity relationships but also guides further structural determination by NMR and X-ray crystallography. The first step in disulfide mapping is the determination of the number of disulfides in a given protein, which can be readily deduced by a simple MS analysis before and after protein reduction. This is nicely illustrated in the ESI MS analysis of recombinant interferon α-2b and GM-CSF, where reduction resulted in a 4 Da shift in the measured M_r, thus indicating the presence of two disulfide bonds [197]. In case of GM-CSF, the ESI mass spectrum prior to and after treatment with β-mercaptoethanol clearly showed a 4 Da shift in the measured M_r (Fig. 4.11 insets), hence confirming the presence of two disulfide bonds in the recombinant protein product.

Fig. 4.11

Positive-ion ESI mass spectrum of recombinant human granulocyte–macrophage colony-stimulating factor (rhGMCSF) in 1 % HCOOH (a) and after treatment with β-mercaptoethanol (b). The deconvoluted spectra are shown in the insets. (Reprinted with permission from Wiley [197])

4.3.2 Disulfide Mapping

Following the determination of the number of disulfide linkages, mapping of the protein’s primary sequence by proteolytic cleavage of the protein between half-cystine residues to produce disulfide-linked peptides and MS analysis of the resulting peptide fragments allows the identification of the existing disulfide arrangement [198]. The potential of MS in this disulfide mapping approach was first realized with the implementation of soft ionization techniques, such as FAB/liquid secondary ion (LSI) [199–201], plasma desorption (PD) [202, 203], and later by the more sensitive method of MALDI [18, 204]. That was nicely illustrated in the disulfide mapping of several therapeutic proteins, such as recombinant human interferon α-2b (INTRON A) [205, 206], human growth hormone [203] and IL-4 [207] by FAB, PD, and MALDI mapping. It should be noted that weak ion signals corresponding to the MH⁺ of the constituent C-containing peptides were also present in the FAB, LSI, PD, and MALDI mass spectra arising from fragmentation of disulfide-linked peptides during the ionization process [208]. This is shown in the LSI mass spectrum of the disulfide-linked tryptic core peptide of rhGM-CSF (expected M_r 7,613) (Fig. 4.12), where additional signals at 5665.2 and 4412.4 Da were also observed due to the presence of the partially reduced peptides T₅-S–S-T₁₁ and T₁₁-S–S-T₁₃, respectively (Fig. 4.12, inset) [197].

Fig. 4.12

Positive-ion Cs⁺ LSI mass spectrum of the HPLC-isolated fraction containing the disulfide-linked tryptic core peptide T₅-T₁₁-T₁₃ of rhGM-CSF. (Reprinted with permission from Wiley [197])

Even though the disulfide-linked peptides yield unique mass spectral signals, the protein fragmentation should be carefully controlled to avoid rearrangement of disulfide bonds (disulfide scrambling), which can take place at neutral and alkaline pH [209]. Therefore, protein cleavage methods performed in aqueous solvents at acidic pH are preferred, such as cyanogen bromide [210] and pepsin [200]. This acidic pH is also optimum for disrupting the protein conformation and making the cleavage sites between half-cystine residues more accessible. That was nicely illustrated in the first report on the disulfide mapping of insulin where FAB MS of peptic digest peptides was combined with Edman analysis for disulfide bond analysis [200]. The intramolecularly linked peptides are identified by the 2 Da increase upon reduction in their constituent half-cystines with β-mercaptoethanol or dithiothreitol, whereas intermolecularly bridged peptides yield protonated MH⁺ signals of the constituent half-cystine-containing peptide fragments.

The advent of ESI [13] has made LC-ESI MS the favorite approach for analyzing the enzyme-generated protein fragments and mapping disulfide linkages in recombinant proteins [198, 206]. Analysis of the peptide mixtures before and after reduction generally allows the identification of the C residues involved in disulfide bonding, taking all aforementioned precautions to minimize disulfide scrambling. It should be noted that ESI MS analysis of disulfide-linked peptides is not conducive to peptide signals arising from partial disulfide bond reduction, as shown in the ESI mass spectrum of the disulfide-linked tryptic peptide T₂₀-T_25,26 of IL-5Rα (Fig. 4.13).

Fig. 4.13

Positive-ion ESI mass spectrum of Sf9 IL-5Rα tryptic fraction (Fig. 4.8, TIC peak 19) containing the disulfide-bonded peptides T₂₀ and T_25,26 with M_r value of 5553

When protein chains are disulfide-linked and proteolysis between half-cystine residues is not possible, identification of the exact location of the disulfide linkage often requires (1) successive proteolytic digestions, such as the ones demonstrated for interleukin-13 (chymotrypsin plus S. aureus V8 protease) [211] and rtPA (Lys-C plus trypsin) [26] or (2) chromatographic separation of the enzyme-derived protein fragments coupled with online MS/MS analysis (e.g., LC-ESI MS/MS), and/or off-line MS/MS analysis and Edman sequencing [212, 213]. This is essential for proteins where three proteolytic fragments are linked by intermolecular disulfides or where two peptide chains contained an intramolecular disulfide and no further proteolysis is possible. The existence of disulfide bonds is usually confirmed by fragmentation of putatively disulfide-linked peptides by MS/MS analysis following ionization by FAB [214], ESI [215], or MALDI PSD [216]. In the MALDI PSD approach, the characteristic ion triplet separated by 33 Da, arising from cleavage at the C–S bond with a concomitant proton transfer [168], can be used as a diagnostic tool for the location and identification of disulfide-paired peptides, even from complex digest mixtures of proteins.

The LC-ESI MS and tandem MS approach is especially valuable in the disulfide mapping of protein receptors and therapeutic proteins having high M_r, such as rtPA and mAb. In mAb, the inter- and intrachain disulfides are responsible for maintaining the characteristic three-dimensional antibody structure, which allows the highly specific antigen binding. Therefore, complete disulfide mapping in mAb is critical for ensuring its therapeutic activity, because incomplete disulfide linkages and/or free sulfhydryl groups can lead to antibody fragments with no antigen-binding activity [217]. In case of the anti-HER2 mAb (Herceptin) that interferes with the HER2/neu receptor and used for the treatment of early-stage breast cancer, the disulfides were completely mapped by LC-ESI MS with the combination of ETD and CID fragmentation [218]. Using ETD cleaves preferentially the disulfides into two polypeptides while CID generates mainly peptide backbone cleavage (with the disulfides intact). This approach was successful in mapping a total of 16 disulfides, 12 intra- and 4 intermolecular, in anti-HER2 mAb and a similar therapeutic mAb. This ETD fragmentation strategy can be further enhanced by CID-MS³ on the dissociated peptides (after ETD) in order to provide corroborating information on the linkage assignment. The same multi-fragmentation approach in combination with multi-enzyme digestion scheme (Lys-C followed by trypsin and Glu-C) was employed in the mapping of the 17 disulfide linkages in human growth hormone [26] and rtPA, as well as for the identification of the unpaired C residue in rtPA [219]. The ETD-MS² spectrum of the disulfide-linked tryptic peptide T₇-T₈-T₉ clearly showed that the unassigned C residue (C83) was found to be paired with either a glutathione or C molecule, which could shed light into the activation or signaling pathway of rtPA. A novel approach based on IM MS was also employed for the rapid characterization of disulfide variants in intact IgG2 mAb [220]. IM MS revealed two to three gas-phase conformer populations for IgG2, compared to only one conformer for IgG1 mAb and a C232S mutant of IgG2, thus indicating that the observed conformers are apparently related to disulfide variants. Therefore, IM MS is a new powerful tool for the characterization of intact mAb and may be useful for fingerprinting higher-order structures of these protein therapeutics.

Finally, disulfide mapping combined with stable isotope-labeling of peptides with ¹⁸O greatly facilitated the identification and characterization of disulfide-linked peptides [221]. Isotope profiles of enzymatically generated peptides produced in 50 % H ¹⁸₂ O (v/v) in H ¹⁶₂ O would produce unique doublets separated by 2 Da, whereas the disulfide-linked peptides should be distinctly different than single-chain peptides [222]. Therefore, the disulfide-linked peptides could be identified in complex peptic digests or chromatographic fractions thereof by MS analysis, and especially MALDI-TOF MS. This procedure is ideally performed in acidic solutions (e.g., peptic digestion) in order to preclude disulfide rearrangement and it may also be used to aid the interpretation of product-ion spectra of disulfide-linked peptides.

4.4 Future Prospects and Challenges

In the past two decades, recombinant protein therapeutics have changed the face of modern medicine as they provide innovative and effective therapies for numerous previously incurable diseases. Protein therapeutics have already a significant role in almost every field of medicine, even though this role is still only in its infancy. The number of recombinant proteins in clinical trials for new and existing therapeutic targets continues to increase annually, as does the total number of protein-based pharmaceuticals reaching the marketplace. The acceptance of the various protein therapies can be attributed to the increasing prevalence of chronic diseases, such as cancer, diabetes, cardiovascular diseases, and neurological/neurodegenerative disorders. In addition, the rising penetration of medical insurance industry has made protein therapeutics available to a wider population. The global protein therapeutics market is expected to grow at an annual rate of 13 % during 2012–2015, arising from the introduction of new protein therapeutics in the major sectors of protein therapeutics market, which include mAb, insulin, interferons, G-CSF, tPA, EPO, coagulation factors, etc.

Recombinant therapeutic proteins for human use must be characterized thoroughly prior to clinical development in order to satisfy the rigorous regulatory requirements (ICH Q6B guidance) [12]. In addition, the manufactured final product should be comparable to that used in preclinical and clinical studies, and its purity, potency, safety, stability, and batch-to-batch consistency should be established. Advances in MS techniques, especially MALDI and ESI, have made MS-based mapping approaches powerful and essential analytical tools for structure characterization of therapeutic proteins and evaluation of recombinant protein heterogeneity including identification of PTMs, sequence variants, and degradation products in recombinant proteins. Structure characterization of all PTMs in a protein is of a great concern for regulatory agencies, such as glycosylation and disulfide linkages. Glycosylation, the most common form of PTM, plays a crucial role in the stability and therapeutic potency of the glycoprotein, as it was demonstrated for rHuEPO. Moreover, changes in levels and types of glycosylation can be associated with certain diseases, such as aggressive breast cancer [223], thus making glycoprotein screening invaluable, not only for diagnostic purposes, but also for design of novel therapeutic drugs. In addition, glycan profiling of normal and diseased forms of a glycoprotein has provided new insights into future research in rheumatoid arthritis, prostate cancer, and congenital disorders of glycosylation [224–226].

In general, LC–MS and tandem MS peptide mapping is the standard and well-accepted approach by the regulatory agencies (FDA, EMA) for identifying PTMs and establishing the recombinant product purity. Nevertheless, a variety of tandem MS experiments should be performed in order to provide insights into the glycan structure (low-energy CID) and peptide backbone sequence/site attachment (ETD and/or high-energy CID) within the same LC–MS run. These MS fragmentation approaches are ideally suited with higher-resolution mass spectrometers, for example, QTOF, IM TOF, and Orbitrap analyzers. The interpretation of the complex and abundant data generated from these experiments undoubtedly requires the support of the growing resources of bioinformatics tools for automated search and identification of glycopeptides and the attached glycans. The advantages of this multi-fragmentation approach (ETD, CID) combined with these high-resolution mass analyzers are also essential in the mapping of disulfide linkages in recombinant protein therapeutics. Even though disulfide linkages are assigned in the initial development stage of the protein, they often need to be reassigned in large-scale production or when the cell production conditions change. Therefore, confirmation of disulfide linkages and identification of any unpaired C location needs to be provided by the aforementioned mapping approach, thus ensuring the proper folding and biological activity of the protein therapeutic product. The latter is especially critical in case of developing innovative treatments using mAbs, which are expected to top the global market in protein therapeutics in the near future. Fast growth in protein therapeutics will also strengthen the emerging segment of bio-generics (biosimilars), which is a key future growth sector due to patent expirations of the branded innovator products. In that case, a thorough characterization of the biosimilar product in terms of glycosylation occupancy and identification of disulfide linkages will be essential for evaluating the comparability between the innovator and biosimilar products. In case of a generic variant of rtPA (TNK-tPA) [150], the analysis strategy was focused on regions that could impact the clot lysis activity such as the glycosylation occupancy at the N184 site and the different extent of oxidation at several M sites. Finally, the advent of more accurate and sensitive instrumentation will enable the development of novel methodologies for the structural characterization of recombinant protein therapeutics and shed some light into the role of specific carbohydrates in many complex biological interactions. That, in turn, will incite the development of novel glycosylated therapeutics for treating infectious, chronic, and other diseases, as well as the improvement of the immunogenicity and pharmacokinetic profiles of existing protein therapeutics.