Intact glycopeptide characterization using mass spectrometry

Abstract

Glycosylation is one of the most prominent and extensively studied protein post-translational modifications. However, traditional proteomic studies at the peptide level (bottom-up) rarely characterize intact glycopeptides (glycosylated peptides without removing glycans), so no glycoprotein heterogeneity information is retained. Intact glycopeptide characterization, on the other hand, provides opportunities to simultaneously elucidate the glycan structure and the glycosylation site needed to reveal the actual biological function of protein glycosylation. Recently, significant improvements have been made in the characterization of intact glycopeptides, ranging from enrichment and separation, mass spectroscopy (MS) detection, to bioinformatics analysis. In this review, we recapitulated currently available intact glycopeptide characterization methods with respect to their advantages and limitations as well as their potential applications.

1. Introduction

Glycosylation, an enzymatic process that attaches glycans to proteins, is one of the most prominent protein posttranslational modifications that play various structural and functional roles in numerous cellular activities[1–5]. The biological function of glycoprotein is closely related to its glycosylation pattern, as demonstrated in the glycosylation of the IgG-Fc moiety[6]. Furthermore, the glycosylation pattern can be altered by various diseases, such as cancer. For example, the glycan pattern of prostate cancer biomarker, prostate specific antigen (PSA), changed significantly between the normal and prostate cancer cells[7]. Enormous efforts have been devoted to identifying glycoprotein disease biomarkers and to determining the glycosylation difference between normal and disease conditions. Up to date, the FDA has approved more than ten different glycoproteins as biomarkers for different types of cancer[8]. In addition to serving as biomarkers for disease prognosis and diagnosis, glycoproteins are also extremely important drug and vaccine candidates. For example, glycosylated monoclonal antibodies are promising drug candidates for diseases such as autoimmunity, inflammation, cancer, and HIV[9–11], and nanoparticles assembled by glycoproteins are developed as vaccines for infectious diseases[12, 13].

Generally, in protein glycosylation, glycans can be conjugated to at least nine different amino acid side chains[14]. The two most common and extensively studied glycosylated proteins are N-linked glycoproteins, in which glycans are covalently attached to the amide group of asparagine (Asn), and O-linked glycoproteins, in which glycans are linked to the hydroxyl group of threonine (Thr) or serine (Ser). The N-linked glycoprotein has a consensus sequence: NX(S/T) where X is any amino acid except proline (Pro)[15]. Structurally, the N-linked glycoprotein has a universal core glycan structure: GlcNAc₂Man₃ (where GlcNAc=N-acetylglucosamine, and Man=Mannose). By contrast, the O-linked glycoprotein has neither consensus sequences nor universal core structures, although GalNAc is the most frequently observed anchor of O-linked glycan. In addition to above-mentioned three types of monosaccharides, more than a dozen different sugars were observed, including galactose, xylose, fucose, 5-N-acetylneuraminic acid, and other kinds of sialic acids.

Similar to other types of proteomics studies, mass spectrometry (MS) based approaches operateing at peptide level have been predominantly applied for glycoprotein characterization, including peptide sequence, glycan composition, and attachment site. However, inherent complexities of intact glycopeptides due to structure heterogeneities present great analytical challenges. First, glycopeptides are often low in abundance, the ion signals of glycopeptides can be suppressed by the majority of co-existing unmodified peptides. Therefore, an efficient enrichment method is necessary. Second, the commonly used collisional based MS fragmentation method like collision-induced dissociation (CID) preferentially dissociates glycan while produces little peptide backbone fragments, which makes identification of peptide sequences difficult. In addition, the low m/z glycan oxonium ions, which are signature ions of glycopeptides[16, 17], may not be detected in the CID MS/MS spectra in ion-trap type MS. Therefore, an alternate fragmentation method is crucial for producing peptide backbone fragments. Third, the lack of sophisticated bioinformatics tools for intact glycopeptide characterization severely hampers glycoprotein research. Even for glycan-only analysis, monosaccharaides attached to amino acids can form a tree structure instead of the linear peptide chain; therefore, new algorithms are needed to accommodate the challenge.

Due to the presented obstacles, the conventional practices for N-linked and O-linked glycopeptide characterization analyze peptide and glycan separately through deglycosylation. Glycopeptides are firstly enriched from complex biological samples, followed by glycan removal through either enzymatic treatment (e. g., PNGase F for N-linked glycoprotein[18, 19]) or chemical reactions[20, 21]. The deglycosylated peptides and/or detached glycans are sequentially collected and analyzed through LC-MS to determine peptide sequences, glycosylation sites, and/or glycan compositions. This approach has successfully demonstrated identification of hundreds of glycosylation sites (N-glycosylation and/or O-glycosylation) in a single analysis [22–24]. Additionally, detailed glycan composition and structure information can be obtained separately [25, 26]; however, this information cannot be directly connected to specific glycosylation sites. An intact glycopeptide strategy (Figure 1) is emerging with encouraging results as an alternative to deglycosylation[17, 27, 28]. This new strategy keeps the glycan attached to its amino acids, which enables to elucidate the glycopeptide sequence, the glycosylation site, as well as the glycan composition simultaneously. In this review paper, we will review current options for each analytical challenge, their pros and cons, as well as applications[17, 27, 29].

Figure 1.

Overall workflow

2. Glycopeptide enrichment and separation

Enzymatic digestion is usually the first sample preparation step before glycopeptide enrichment and MS characterization. Trypsin is the most commonly used enzyme in glycoprotein digestion because it specifically cleaves after arginine and lysine, and produces more specific and relatively long glycopeptides for better glycosylation site localizations and more confident peptide identifications. However, larger glycopeptides are often more difficult to be ionized and detected in the routine analysis. Additionally, the cleavage site may be covered by the massive oligosaccharides, which results in miss-cleavage and long and even larger size glycopeptides. On the other hand, non-specific enzymes, such as Pronase E or Protenase K, have been applied to generate smaller size glycopeptides with four to six amino acids[30, 31] for better detection in MS. However, many of these small nontryptic glycopeptides are not unique peptides and additional information is needed to confidently assign the glycosylated sites. In addition, the fragmentation patterns of nontryptic glycopeptides need to be better understood.

Because of the often-low abundance of glycopeptides in the digestion mixture, the enrichment step is the prerequisite of glycopeptide characterization. Many enrichment and separation methods have been developed for general or specific types of glycopeptides. Here, we will review various enrichment and separation strategies including lectin affinity, hydrazide chemistry, affinity enrichment, hydrophilic interaction liquid chromatography (HILIC), porous graphitized carbon (PGC) separation, and ion mobility spectrometer (IMS) separation. General peptide separation approaches such as reverse phase liquid chromatography (RPLC) and gel separation are not included in this review. However, to improve the enrichment and separation efficiency, orthogonal two-dimensional enrichment/separation composed of glycopeptide specific approaches only or with general peptide separation approaches are often applied[17, 27].

2.1 Lectin affinity

Lectins are a group of non-enzymatic proteins derived from plants, animals, or microbes, which recognize and bind carbohydrates[32, 33]. More than 60 lectins are commercially available[34]. The specificities of the commonly used lectins were described in detail in other review papers[32, 34]. Some lectins, including wheat germ agglutinin (WGA), concanavalin A (Con A), and jacalin (JAC), recognize a wide range of glycan structures, so they can be applied for the enrichment of larger portion of glycoproteome. Others, including Sambucus nigra agglutinin (SNA) and Maackia amurensis leukoagglutinin (MAL), have narrower specificity, so they can be used for the enrichment of a small subset of the glycoproteome[32, 34]. Due to the binding specificity of lectins, single[35], serial[36], or multi-lectin[37] affinity chromatography can be used for glycoprotein purification. Another approach so called SimpleCell utilize a universal and stable generic engineering strategy to block the elongation of O-glycans. Combining lectin enrichment and ETD/HCD based mass spectrometry analysis, Steentoft identified 600 O-glycoproteins with about 3000 O-glycosylation sites on these engineered human cells[38].

However, none of the available lectins, or even the combination of different lectins, can capture the entire glycoproteome, so some glycans might be lost during the lectin enrichment[34, 39]. The efficiency of lectin enrichment is also contingent on lectin's accessibility to the target glycan within glycoproteins[32]. In addition, since the concentrated saccharide elution media is incompatible with MS, lectin affinity chromatography cannot be coupled with MS directly.

2.2 Other affinity enrichment approaches

Analysis of sialylated glycoproteins is of particular interest, partly due to the association of this type of glycoproteins with some diseases[40–42]. Titanium dioxide (TiO₂) enrichment is a method highly efficient for glycopeptides containing sialic acids as terminal sugars[43]. Under highly acidic conditions, both carboxylic acid and hydroxyl group of sialic acids interact with TiO₂ through multipoint binding, while non-modified peptides or neutral glycopeptides exhibit no affinity to TiO₂[43]. This method has been successfully applied to explore human plasma and saliva sialiomes: nearly 200 glycopeptides and 100 glycosylation sites were identified[43]. A detailed protocol is also available[44]. The drawback of this strategy is that elution of oligo and polysialylated species is difficult [45].

Azido sugar metabolic labeling is another type of affinity enrichment, which is particularly efficient for enriching low abundance glycoproteins in secretome with the presence of serum[46]. Azido sugars are incorporated into glycan structures by using existing biosynthetic pathways of mammalian cells, and the azide moiety can be utilized to enrich glycoprotein or glycopeptides through click-chemistry and affinity purification. The drawbacks of this method are the time-consuming sugar incorporation process, and many types of samples cannot be obtained through culture processes.

2.3 Hydrazide chemistry

The principle of hydrazide chemistry enrichment is that the cis-diol groups of glycan residues are oxidized by sodium periodate to form aldehydes, the oxidized glycans are captured using hydrazide beads, and then the attached N-linked glycoproteins/glycopeptides are released by enzymatic deglycosylation using PNGase F[47]. The application of hydrazide chemistry is primarily limited in N-linked glycoprotein analysis, due to the lack of an efficient approach to cleave O-linked glycoproteins and glycopeptides[47, 48]. Furthermore, this enrichment approach only delivers deglycosylated glycoproteins and glycopeptides, but not intact glycoproteins and glycopeptides. Recently, this approach was extended to characterize intact sialylated glycopeptides. The selectively oxidized sialylated glycoproteins were covalently bound to hydrazide beads and were released by acid hydrolysis with the selective removal of the sialic acids at the terminal positions of glycan chains of the enriched glycopeptides[49, 50].

2.4 Hydrophilic interaction liquid chromatography (HILIC)

HILIC has been employed extensively to enrich and separate carbohydrates[51]. In general, the samples are loaded in the non-polar mobile phase, usually acetonitrile, and analytes are retained to the polar stationary phase through hydrogen bonding, ionic interactions, and dipole-dipole interactions, followed by elution by increasing the water content in the mobile phase[51]. Among the numerous types of HILIC stationary phases, amide and zwitterionic (ZIC-HILIC) are particularly useful to enrich and separate glycopeptides[17, 27, 39, 52, 53]. We have recently demonstrated that the glycan enrichment using ZIC-HILIC followed by dialysis to remove non-specifically bound hydrophilic small molecules (e.g., Lys, and Arg) is a simple, fast, and efficient approach for the downstream NMR analysis of N- and O-linked glycans[54].

HILIC coupled with MS is a versatile technique for glycomics analysis[51]. HILIC performs excellently in enriching glycopeptides from non-glycosylated peptides[17, 27, 39]. Furthermore, HILIC separates N-linked and O-linked glycopeptides as well as isomeric glycopeptides[51–53]. By using MS compatible elution buffers, HILIC can be on-line coupled to nano-electrospray MS, which allows direct analysis of the enriched glycopeptides[52, 53]. The disadvantage of HILIC enrichment is that the solubility of analytes is lower in organic mobile phase [45].

2.5 Porous graphitized carbon (PGC)

The PGC stationary phase is composed of pure graphitized carbon, which is highly resistant to extreme LC conditions (e.g. pH, temperature) and thus is advantageous over the conventional silica-based phase[55–57]. PGC has long been adopted to enrich and separate glycan and glycopeptides[17, 58–61]. The retention mechanism of polar analytes by PGC, although not well understood, is drastically different from that of conventional nonpolar phases. Two major explanations exist for the polar retention effect on graphite: dipole-dipole interaction induced by charged analytes, or electronic repartition on the graphite surface[62].

An and colleagues[61] pioneered using PGC solid phase extraction (SPE) to enrich non-specifically digested glycopeptides. In addition to carbohydrate enrichment, PGC has unique strength in separating glycan isomers[58–61], since the retention of analytes on graphitized carbon does not only depend on its hydrophobicity, but also on the interaction of polarized groups with the graphite. The interaction strength is determined by the type of interaction and interaction area. Stronger interaction or larger interaction area results in better retention. Because of PGC’s consistent performance and efficient separation of both neutral and anionic oligosaccharides, PGC is also widely used for oligosaccharide separation [62].

2.6 Ion mobility spectrometer (IMS) separation

Isomeric monosaccharides, such as GlcNAc and GalNAc, are a challenge for MS-based glycan structure reconstruction because these isomers have the same atomic compositions and hence have identical masses. The employment of additional gas-phase separation dimension using IMS, provides an avenue to overcome the limitation, because ionized molecules with identical mass but different structures can be separated by IMS according to their collisional cross section size[63]. It has been demonstrated that the GlcNAc and GalNAc oxonium ions were partially separated by IMS, and the same glycopeptides with either GlcNAc or GalNAc attachments were detected with different collisional cross section size[64]. Furthermore, glycan isomers with different linkage or branching styles showed different drift time patterns, which provided insightful information to resolve glycan structures[65]. With the development of IMS technique and availability of commercialized instruments, IMS application for carbohydrate isomers separation and glycan structure characterization is increasing significantly [64–71]. However, the software development for IMS data analysis is currently behind the instrument development, which hinders the application and utilization of IMS.

Overall, affinity based enrichment approaches such as lectin or TiO₂ enrichment are highly specific, and can be applied to enrich low concentration glycoproteins or peptides in a complex sample such as serum or secretome. However, many of these affinity interactions are specific to certain carbohydrate structures, making them impractical for global glycopeptide analysis. On the other hand, HILIC or PGC enrichments are more generic because they utilize polar retention of glycopeptides on the column; however, these methods have limited distinguishing power between hydrophilic peptides and glycopeptides. Additional separation dimensions such as IMS can be directly coupled on-line with MS for glycopeptide isomer separations and glycan structure characterization.

3. MS Fragmentation

A thorough elucidation of intact glycopeptides requires detailed information from both the peptide backbone and its associated monosaccharaide. In this section, we will discuss multiple tandem MS (MS/MS) strategies that have been developed and applied to characterize intact glycopeptides, including collision induced dissociation (CID), electron transfer dissociation (ETD), and higher energy collisional dissociation (HCD) (Figure 2 as an example of the comparison of different fragmentation approaches on a glycopeptide[27]). Various fragmentation approaches can be combined either sequentially (i.e., MS3) or in parallel (i.e., alternating) to provide more comprehensive information.

Figure 2.

Comparison of CID-, HCD-, and ETD-MS for fragmentation of a C. jejuni PEB3 glycopeptide.[27]

3.1 CID

CID is the earliest and most prevalently utilized tandem MS approach to generate diagnostic fragment ions. CID breaks the CO-NH bonds that produce b (N-terminal fragment) and y (C-terminal fragment) ions of non-modified peptides. However, glycosidic linkages of glycopeptide are preferentially broken to produce carbohydrate Y series ions because they are more labile than the amide bonds (nomenclature for describing the fragmentation of carbohydrates as in Figure 3[72]), which reveal predominantly the glycan moiety structure. Due to this limitation, CID is seldom applied alone to study intact glycopeptides. Nonetheless, by combining CID produced Y series ions and high mass measurement accuracy on short glycopeptide, Nwosu et al., (2011) identified more than 200 N- and O-linked glycopeptides corresponding to 18 glycosites from a mixture of glycoprotein standards.

Figure 3.

Nomenclature for describing the MS fragmentation of carbohydrates. In this nomenclature, ions retaining the charge on the non-reducing terminus are named A, B and C, and the ions retaining charge on the reducing terminus are X, Y and Z. A and X correspond to cross-ring cleavages, whereas B, C, Y and Z correspond to glycosidic cleavages. Subscript numbers denote the cleavage position, starting at the reducing terminus for the X, Y and Z ions, and at the non-reducing terminus for the others. In the case of ring cleavages, superscript numbers are given to show the cleaved bonds.[72]

An alternative approach is adding one more fragmentation stage (MS³)[49, 50, 53, 73]. In this approach, the selected Y ion generated by MS/MS was subjected to a second ion fragmentation cycle to generate b and y ions from a peptide backbone[49, 50, 73]. Halim et al., (2012) applied this approach in combination with electron capture dissociation (ECD) fragmentation to study urine glycoproteins, and identified more than 100 N- and O-linked glycopeptides. However, this approach is limited by the requirement of sufficient parent ion intensity to generate useful information from the MS³ experiment. Moreover, currently there is no available software to automatically interpret and combine information from different fragmentation stages, which further limited the applicability of the MS³ approach, particularly for complex biological samples.

3.2 Electron transfer dissociation (ETD)

ETD is an alternative MS fragmentation method and is particularly effective for glycopeptide. During ETD fragmentation, a multiply protonated peptide receives an electron from a radical anion and cleaves at the N-C_α bond to produces c (N-terminal fragment) and z (C-terminal fragment) ions[74]. In contrast to CID, labile post-translational modifications, such as glycosylation, are generally retained on the peptide backbone and are largely unaffected by the ETD fragmentation process.

ETD can be used alone for glycopeptide characterization[28, 75, 76]. Trinidad et al., (2013) conducted a study of N- and O-linked glycoproteins in mouse synaptosome. By applying ETD to lectin enriched glycopeptides, over 2500 unique N- and O-linked glycopeptides on 453 proteins in murine synaptosome were identified. However, since ETD does not cleave glycosidic linkages effectively, only monoisotopic masses of oligosaccharides were obtained, and no information of the monosaccharide units was provided.

More frequently, ETD is utilized in combination with complimentary fragmentation approaches, like CID, to collect fragments from both the peptide backbone and attached glycan [27, 77]. Alley et al., (2009) demonstrated that the CID/ETD alternative fragmentation is capable of obtaining comprehensive information of the entire glycopeptide from model glycoproteins. The major drawback of ETD is its relatively narrow useful m/z range [27, 75, 77]. The fragmentation efficiency of ETD is relatively low (~20%) [77], especially for precursors with m/z higher than 850[75]. This disadvantage is critical for trace amount of analytes in biological samples. In addition, ETD fragmentation works most efficiently on glycopeptides with single HexNAc[78] or GalNAc[38, 79].

3.3 Higher energy collisional dissociation (HCD)

HCD, a newly developed dissociation approach which is unique to Orbitrap MS by Thermo Scientific, also produces b and y type fragment ions. Compared with CID, HCD has several advantages. First, HCD is performed in a dedicated octopole collision cell at the end of the “C-trap”, which enhances trap efficiency in the low m/z range, while CID fragments are often limited by the “one-third” m/z cutoff rule[80]. Unlike CID, small diagnostic oxonium ions of monosaccharaides are detectable in HCD[16, 17, 27]. Second, it efficiently dissociates the peptide backbone while preserving the attached labile modifications[27, 81, 82].

Many studies have demonstrated that the HCD did provide beneficial information for glycopeptide characterization[17, 27, 29, 82, 83]. Segu and Mechref (2010) [82] first noticed that Y1 ions (peptide + GlcNAc) of N-linked glycopeptides were the most abundant ions in the HCD MS/MS spectra. The Y1 ion was selected and subjected to a second HCD fragmentation to collect b and y ions from the peptide backbone. Cao et al., (2014) [17] further examined different fragmentation pattern of N- and O-linked glycopeptide according to different normalized collisional energy (NCE) value. Their result suggested that HCD with lower NCE values preferentially fragmented the glycan, while HCD with higher NCE values preferentially fragmented the peptide backbone. By applying HCD with alternating low and high NCE values, comprehensive structure information of both the sugar chain and the amino acid chain were collected.

In many applications, HCD is also combined with CID or ETD for complimentary fragmentation[27, 29, 84]. Scott et al., (2011) [27] applied both CID/HCD and CID/ETD to analyze N-linked glycoproteome of bacteria C. jejuni. Compared with CID/ETD, CID/HCD has a higher duty cycle, since ions are split between the ion trap and C-trap for CID/HCD, while delay time is needed for the switch between capturing ions for CID and ETD[27]. HCD/ETD is another combination which showed promising results[29, 84]. Yin et al., (2013) [29] applied and compared alternating HCD-ETD and HCD-product-dependent ETD methods to study endothelial cell secretome. The two methods selected different sets of glycopeptides for fragmentation; therefore, the identified glycopeptides had little overlap. Overall, complimentary fragmentation in general provides greater confidence in the identification of glycopeptides than studies relying on a single approach.

HCD and higher stage CID fragmentation were also applied to provide glycan structure information[85]. Halim et al., (2014) demonstrated that oxonium ion spectrum generated using HCD or CID MS3/MS4 were distinctly different between GlcNAc and GalNAc or among glycans with the same monosaccharide composition but different linkages[85].

4. Bioinformatics

Due to the complexity of glycoprotein, many different combinations of glycan and peptide composition can be isobaric. Based only on intact mass, early software like GlycoMod[86] yields multiple compositional possibilities. To determine the correct composition, tandem MS (MS/MS) is often performed in addition to MS. However, fragmentation increases the complexity of the data and makes the analytical task more challenging. Manual analysis of each tandem spectrum is laborious, time-consuming and requires significant expertise. Recently, a number of algorithms have been developed with automated approaches for glycopeptide spectra interpretation, including SimGlycan[87], GlypID 2.0 /GlycoFragWork[16, 88], ByOnic[89], GlycoPep Grader[90], SweetSEQer[91], GP Finder[92], Sweet-Heart[93], GPQuest[94], GlycoFinder[17], GlycopeptideSearch (GPS)[95], and GlycoMasterDB[96]. The attributes of each software are summarized in Table 1.

Table 1.

Comparison of the currently available bioinformatics tools for glycopeptide analysis

Software	Description	Availability	Reference
GlycoMod	Uses intact MS without tandem MS data; searches against known peptide sequences for both N- and O-linked glycopeptides, and reports all results without confident scores	Non- commercial; available through downloading	Cooper, et al., 2001[86]
SimGlycan	Raw/standard MS data files input, MS/MS database searching for glycans, scoring and ranking glycan results	commercial	Apte&Meitei, 2010[87]
GlypID 2.0/ GlycoFragWork	MS and MS/MS input, raw/mzXML standard format. Use CID, ETD, and HCD spectra. Scoring available and false discovery rate (FDR) available in later version. Only applicable to N-linked glycopeptide.	non- commercial; available through downloading	Mayampurath et al., 2011[16]; Mayampurath et al., 2014[88];
ByOnic	Uses tandem mass spectra (MS/MS) spectra, both N- and O- linked glycopeptides, does not require known glycans, supports for all types of fragmentation	Commercial	Bern et al., 2012[89]
GlycoPep Grader	Uses intact MS and tandem MS, designed for N-linked glycopeptide, needs protein sequence, does not require known glycans, MS/MS analysis in charge state dependent fashion, works on low resolution CID data	Non- commercial; available through downloading	Woodin, et al., 2012[90]
SweetSEQer	de novo analysis of tandem MS for glycoconjugates, can be customized and added to other software	Open-source, non- commercial; available through downloadinga	Serang et al., 2013[91]
GP Finder	Uses intact MS and tandem MS, needs protein sequence, needs known glycans, designed for both N- and O-linked glycopeptide, independent of the type of enzyme. An improved version of GlycoX	non- commercial; available through downloadingb	Strum et al., 2013[92]
Sweet-Heart	Uses MS2 data to predict glycopeptides and design targeted MS3 sequence. Information from MS2 and MS3 experiments integrated to derive peptide backbone sequence and glycoform. Applicable to N- linked glycopeptides only.	Non- commercial	Wu et al., 2013[93]
GPQuest	MS and MS/MS input, HCD spectra. Uses glycan oxonium ions to select intact glycopeptide spectra, then compares the detected spectra to the theoretical glycopeptide spectra library to identify N-linked glycopeptides.	Non- commercial;	Eshghi et al., 2015[94]
GlycoFinder	Specifically designed for tandem MS (MS/MS) spectra with alternating HCD NCE parameters. Uses glycan oxonium ions to select candidate glycopeptide spectra. HCD spectra obtained with low NCE values are analyzed in a similar manner as de novo sequencing for possible monosaccharide composition; HCD spectra obtained with high NCE values are analyzed by an approach similar to USTags[98] and manually validated for peptide sequence.	Non- commercialc	Cao et al., 2014[17]
GlycoPeptideSearch (GPS)	Checks for oxonium ions in CID MS/MS spectra to select glycopeptide spectra. Each spectrum is checked for intact- peptide fragment ions and followed by glycan structure matches. Peptide-glycan match in each spectrum is evaluated and glycopeptide FDR is available. Designed for N-linked glycopeptides.	Non- commercial; available through downloading	Chandler et al., 2013[95]
GlycoMaster DB	Uses HCD-only or HCD/ETD MS/MS spectra. Checks for oxonium ions and ion ladders in HCD spectra to select glycopeptide spectra and generates glycan structure based on HCD fragments. Identifies peptide backbone of glycopeptide in ETD spectra. When ETD data is not available, peptide sequence is identified by matching the calculated peptide mass and considering the glycopeptide sequon. Designed for N-linked glycopeptides. Matching scores for glycan and peptide are calculated.	Non- commercial; available through downloading	He et al., 2014[96]

Notes:

^aavailable at the following website: http://software.steenlab.org/

^bLink changed to http://lebrilla.faculty.ucdavis.edu/links/glycopeptide-finder/

^cavailable upon request

With large amounts of nonglycosylated peptides co-produced during protease digestion, glycopeptides are often suppressed or not selected for fragmentation. Using nonspecific proteases[92] or performing glycopeptide enrichment prior to testing are two approaches to circumvent this issue[97]. Recently, a software enrichment strategy on accurate mass measurement and mass defects was developed to quickly discriminate between peptide and N-glycopeptide signals in mass spectrometry. The method based on intrinsic mass defects of elements can be applied to diverse glycoproteomic problems without the need for prior knowledge of protein sequence and glycans[97].

5. Expert commentary and Five-year view

The field of intact glycopeptide study is growing substantially, due to advancements in sample preparation methodologies, instrumentation, and specialized software. Recent studies reported confident identification of hundreds to thousands intact glycopeptides from complex biological samples[27, 28, 94, 99]. However, the identified glycopeptides still only comprise a small percentage of the whole glycoproteome. However, there are remaining challenges for intact glycoprotein analysis. (1) There is the lack of high throughput, universal intact glycopeptide enrichment techniques. As we discussed earlier, affinity based enrichment approaches are only specific to a subclass of glycan structures, while hydrophilic-interaction based separation approaches cannot really distinguish glycopeptides with hydrophilic nonglycopeptides. To address this issue, a multiplexed affinity purification approach such as combined schemes of affinity purifications can be incorporated to improve the throughput. An attractive direction is to develop a chemical reaction based Azido sugar labeling approach to combine multiple labeling reagents for various glycan structure groups. After labeling, a single step click-chemistry as well as the following affinity purification can be performed for the following MS detection. In addition, despite of the recent advancements and improvements of the intact glycopeptide purification and separation, the intact glycopeptide sample preparation workflow is time consuming, and there is a critical need for developing automation approaches in both sample preparation and data analysis step. (2) There are no efficient glycan structure elucidation techniques. Glycan structure annotations using current fragmentation approaches in MS often rely on oxonium ions and neutral mass loss. It is very difficult to distinguish structure isomers, or define monosaccharides with the same mass (i.e., mannose and glucose). Recently, several studies based on the oxonium ion spectrum produced by different collision methods and IMS suggested improved isomeric glycan separation and provided more information about glycan structure elucidation[64, 65] .This review restricted discussion in qualitative characterization of glycoprotein, however, quantitative description is another crucial challenge of characterization. With rapid technology improvement, these issues are expected to be resolved in the near future.

Key issues.

• A universal approach to enrich both O-linked glycopeptides and N-linked glycopeptides needs to be established to improve the coverage of glycoproteome.

• Improved mass spectrometry fragmentation approaches should be developed to confidently assign both glycosylation site localization and glycopeptide sequencing.

• Improved bioinformatics tools are needed to confidently and automatically assign intact glycopeptides with confidence.

• Intact glycopeptide isomers are difficult to separate using current available separating techniques.

• Strategies for quantitative glycopeptide analysis are still to be developed.