Chapter 50Structural Analysis of Glycans

Glycoproteins

A glycosylated protein typically presents one or more diffuse bands during gel electrophoresis, resulting from heterogeneity of the attached glycans. Even when visualized by protein staining reagents, this phenomenon is often the first indication of the presence of glycans. Some glycoconjugates of very high molecular weight, such as mucins and proteoglycans, do not enter ordinary gels or, if they do, they migrate as heterogeneous smears. Agarose gels or combination polyacrylamide-agarose gels may be useful in this situation. Several analytical options are available to investigate the presence of glycans further (e.g., the PAS stain described above). Treatment of glycoproteins with endoglycosidases (e.g., peptide-N-glycosidase F [PNGase F], endoglycosidase F2 [Endo F2], endoglycosidase H [Endo H]; see Table 50.1 and Figure 50.1) is another option. If this results in a mobility change of one or more of the bands on the gel, the presence of N-glycans is indicated. O-glycanase (endo-α-N-acetylgalactosaminidase; Table 50.1) can be used for the specific identification of O-glycans, sometimes requiring pretreatment with other enzymes to expose the disaccharide core. Removal of individual sugars by exoglycosidases such as sialidase or β-galactosidase (Figure 50.1) may also result in a mobility change depending on the number of residues removed. However, some glycans cannot be detected by these treatments as they are resistant to the enzymes used. Resistance can result from modifications to glycan hydroxyl groups (e.g., sulfation, acetylation, or phosphorylation; Chapter 2), glycosidic linkages that are not recognized by the enzymes, or steric inaccessibility of the glycan. Complete removal of N- and O-glycans can be achieved by chemical treatments (e.g., hydrazinolysis or β-elimination) but peptide damage usually precludes further analysis by gel electrophoresis. The glycan may also be partially degraded (e.g., by loss of O-acetylation).

FIGURE 50.1.

Glycosidases used for structural analysis. (Left) A biantennary N-glycan is shown with exoglycosidases that remove each monosaccharide sequentially. Exoglycosidases act only on terminal sugars. Also shown are endoglycosidases that remove an intact N-glycan. (more...)

TABLE 50.1.

Table of enzymes for glycan analysis

Proteoglycans

Proteoglycans (Chapter 17) may be separated by agarose gel electrophoresis and by ion-exchange chromatography, which separates on the basis of charge conferred by sulfate groups. Treatment of proteoglycans with GAG lyases (Table 50.1) will produce a shift in mobility on a gel, condensing the proteoglycan smear into discrete bands. After removal of much of the glycan portion, antibodies that recognize the remaining “stub” glycans may be used in western analysis.

Glycolipids

Typically, the analysis of glycolipid glycans by NMR or MS is preceded by their purification using chromatographic methods. Mixtures of glycolipids can be fractionated by TLC, and staining of TLC plates with glycan-reactive reagents may allow detection of individual glycolipids. Using different reagents, it is possible to recognize gangliosides (e.g., resorcinol-HCl detects Sia) or neutral monosaccharides (e.g., orcinol-sulfuric acid detects all monosaccharides) in a TLC band. Reagents are also available for detection of sulfate and phosphate groups on glycolipids. Some prepurification of the crude extract is usually preferred (e.g., Folch partitioning and ion-exchange chromatography). These procedures separate nonpolar or nonionic lipids from polar lipids (e.g., glycosphingolipids; Chapter 11) and those that contain charged groups (i.e., gangliosides, phospholipids, and sulfatides). It is also common practice to deduce the presence of specific sugars by evaluating the shifts produced in the migration position of a band following a chemical or enzymatic treatment. Glycolipids on TLC plates can also be detected by monoclonal antibodies, lectins, or even intact microorganisms expressing glycan-specific receptors that recognize specific glycan features (Chapter 48). Detailed structural features may be identified by running the TLC in a second dimension following a specific treatment. On a larger scale, glycolipids are separated using column chromatography or by high-performance TLC on silica plates.

GPI Anchors

GPI-anchored proteins (Chapter 12), with their lipid, protein, and glycan moieties, have unique physicochemical properties that can be exploited for their detection. The nonionic detergent Triton X-114 at low temperature (4°C) extracts soluble and integral membrane proteins as well as GPI-anchored proteins. When the solution is warmed, two phases separate, and GPI-anchored and other amphiphilic proteins remain associated with the detergent-enriched phase. GPI-specific phospholipases can be used to cleave GPI anchors for further characterization. Successful cleavage by GPI-specific phospholipases can be assessed by subsequently analyzing samples by SDS-PAGE, because removal of the GPI anchor causes a shift in molecular mass. This is a common diagnostic method for identifying the presence of a GPI anchor on a protein of interest. Another method is to treat the GPI-anchored protein with nitrous acid, which cleaves the unsubstituted glucosamine residue that links the glycan to the phosphatidylinositol.

Plant and Bacterial Polysaccharides

This class of glycans contains many structures, including homo- and heteropolysaccharides, neutral and ionic polysaccharides, and linear and branched structures, with widespread molecular sizes ranging from a few monosaccharide units to thousands (Chapters 3, 21, and 22). These polysaccharides are typically extracted with water, salts, chaotropic agents, or detergents and isolated by precipitation with alcohols. Detection is based on refractive index (RI) or colorimetric reactions, because sample quantity is not usually a limitation.

Release of Glycans from Glycoconjugates

When glycans are released before structural analysis, it is best to use a quantitative release procedure that neither destroys nor alters the glycan. Ideally, information regarding the type of linkage between the glycan and its liberated protein or lipid should be retained, although this is not always possible.

Glycolipids can often be isolated and characterized by MS and/or NMR without the need for release of glycans, but if necessary enzymatic methods can be used, or for glycosphingolipids, ozonolysis will separate lipid from glycan.

Complex, hybrid and oligomannose N-glycans can be released from glycoproteins with N-glycosidases such as PNGase F or PNGase A. Endo H may be used for the selective release of high-mannose and hybrid N-glycans, but complex N-glycans are resistant (Figure 50.1).

Chemical approaches suitable for release of glycans from a protein include hydrazinolysis, which releases N-glycans and/or O-glycans, depending on the experimental conditions. Strong base treatment can, under carefully controlled conditions, release only O-glycans in a process called β-elimination; it is generally accompanied by reduction with borohydride to give an alditol. More recently methods have been developed in which base treatment is accompanied by derivatization with pyrazolone, which acts as a UV-absorbing label during chromatographic separation. However, all of the chemical methods can result in partial or complete loss of labile modifications, such as O-acetylation or sulfation.

Separation of Glycans

Glycans released from a glycoconjugate usually form a complex mixture. Even when only one glycosylation site in the protein is occupied, individual molecules can bear different glycan species generating multiple glycoforms. The high-throughput analysis of these glycan mixtures using glycomics technologies is described in Chapter 51.

Preparation of individual glycan samples for structural analysis, whatever their source, will rely on separation techniques. Chromatographic separations commonly used to isolate pure glycans, or at least glycan mixtures of reduced complexity, include size exclusion chromatography (SEC), strong or weak anion exchange chromatography (SAX or WAX), and some forms of reverse-phase high-pressure liquid chromatography (HPLC). Electrophoretic methods include capillary electrophoresis (CE) and fluorophore-assisted carbohydrate electrophoresis (FACE).

HPLC-based separation is feasible, even for large samples (>10 mg), if the mixture contains fewer than ∼50 glycan species. Individual fractions can be recovered for further structural analysis by MS or NMR. Once liberated from their glycoconjugates, glycans with free-reducing termini (Chapter 2) can be chemically labeled with fluorescent tags, such as 2-amino-pyridine (2-AP) or 2-AB, providing detection sensitivity that rivals the sensitivity achieved with radiolabels. Advantages of this method include more facile purification of the labeled glycans and a wider variety of options for chromatographic separations and structural analysis techniques. Nonlabeled glycans may be detected by (relatively insensitive) RI methods, or by more sensitive pulsed amperometric detection (PAD).

If a label is introduced at the reducing end, it is possible to use sequential exoglycosidase treatments (Figure 50.1) and look for shifts of the oligosaccharide in a chromatographic system (e.g., paper chromatography, HPLC, TLC) as an indication of susceptibility to the enzyme. Comparison with known standards treated in the same manner allows tentative glycan identification. However, well-characterized standards of known structure are difficult to obtain in a pure form and there are nearly always “contaminating” peaks in a chromatogram that cannot be readily identified. It is very important to note that separation profiling should not be confused with actual structural analysis, because coelution with a standard could occur with a variety of different structures.

Determination of Linkage Positions

Linkage analysis is a well-established and ingenious approach for determining linkage positions. The principle of this method is to introduce a stable substituent (an ether-linked methyl group) onto each free hydroxyl group of the native glycan. Glycosidic linkages are then cleaved by acid hydrolysis, producing partially methylated monosaccharides with free hydroxyl groups at the positions that were previously involved in a linkage. The partially methylated monosaccharides are ring-opened with a reducing agent (normally borodeuteride) to introduce a new hydroxyl group and, more importantly, a deuterium atom at C-1, which helps identify the reducing end of each monosaccharide. All the free hydroxyl groups are then acetylated resulting in partially methylated alditol acetates (PMAAs) that can be identified by a combination of GC retention times and electron impact (EI)-MS (Figure 50.2). The masses of fragments produced by impact of high-energy electrons on PMAAs identify substitution sites in some cases, but fragmentation patterns of similarly substituted isomeric monosaccharides (e.g., Glc and Gal) can be nearly identical. Thus, definitive identification of monosaccharides requires, in addition to the analysis of the MS pattern, the comparison of GC retention times with those of known standards (e.g., all peracetylated 2,3,4-tri-O-methyl-hexoses produce the same EI-MS spectrum, but peracetylated 2,3,4-tri-O-methylgalactitol elutes later than peracetylated 2,3,4-tri-O-methylglucitol). This type of analysis identifies terminal residues (they are methylated at every position except the hydroxyl group at C-1 and the ring oxygen), indicates how each monosaccharide is substituted, including the occurrence of linkage and branching points, and allows the determination of the ring size (pyranose [p] or furanose [f]) for each monosaccharide. However, linkage analysis provides no sequence information and cannot determine the α- or β-anomeric configuration.

FIGURE 50.2.

An example of linkage analysis showing a bacterial O-linked branched hexasaccharide with a sequence of Rha1-3Glc1-(Glc1-3GlcNAc1-)2,6Glc1-6GlcNAc. The O-glycan is reductively eliminated from proteins before successive steps of permethylation, acid hydrolysis, (more...)

Mass Spectrometry

EI-MS in monosaccharide composition and linkage analyses is covered above. In this section, three other types of ionization used in MS—fast atom bombardment (FAB), matrix-assisted laser desorption/ionization (MALDI), and electrospray ionization (ESI)—are described. All three technologies permit the direct ionization of nonvolatile substances, and are applicable to intact glycoconjugates, as well as fragments thereof. FAB-MS was the first to be developed and was the premier MS method for sequencing glycans and glycoconjugates through the 1980s and 1990s. Although FAB-MS has now been largely replaced by ESI-MS and MALDI-MS (which offer greater sensitivity and applicability to high molecular weight glycans and glycoconjugates), the latter techniques continue to exploit sample handling strategies originally developed for FAB-MS. In addition, the knowledge of glycan fragmentation pathways that emerged from the FAB-MS research of the 1980s underpins today's ESI- and MALDI-MS analyses. Among the structural features that can be defined by MS methods are (1) degree of heterogeneity and type of glycosylation (e.g., N-glycan vs. O-glycan; oligomannose, hybrid, or complex N-glycans); (2) sites of glycosylation and identity of the protein/lipid carrier; (3) glycan branching; (4) the number and lengths of antennae, their composition, and substitution with Fuc, Sia, or other capping groups such as sulfate, phosphate, or acetyl esters; and (5) complete sequences of individual glycans.

In the FAB-MS experiment, samples are dissolved in a liquid matrix and ionization/desorption is effected by a high-energy beam of particles fired from an atom or ion gun. In MALDI-MS experiments, the sample is dried on a metal target in the presence of a light-absorbing matrix until matrix crystals containing trapped sample molecules are formed. Ionization of the sample is effected by energy transfer from matrix molecules that have absorbed energy from laser pulses. MALDI-MS is the preeminent technique for screening for molecular ions in complex mixtures and it is a powerful tool for glycomic profiling (Chapter 51). For ESI-MS, a stream of liquid containing the sample enters the source through a capillary interface, where the sample molecules are stripped of solvent, leaving them as multiply charged species. ESI-MS can be coupled to nano- or capillary-bore LC permitting online LC/ESI-MS analysis. This method is particularly useful when complex mixtures of peptides and glycopeptides are being examined (e.g., after proteolytic digestion of a glycoprotein).

In principle, MS provides two types of structural information—the masses of intact molecules (the molecular ions) and the masses of fragment ions. However, in both ESI- and MALDI-MS, the ionization process is usually not sufficiently energetic for fragment ion formation. To overcome this, ESI and MALDI instruments normally have two analyzers in tandem, which allows detection by the second analyzer, of fragment ions produced after molecular ions selected by the first analyzer undergo collisions with an inert gas in a chamber placed between the two analyzers. This is referred to as collision-activated dissociation (CAD) or collision-induced dissociation (CID). MALDI-MS/MS is commonly performed on instruments that have two time-of-flight (TOF) analyzers in tandem (TOF-TOF), whereas the most popular instruments for ESI-MS/MS have either a quadrupole (Q) as the first analyzer and an orthogonal TOF as the second analyzer (QTOF) or a linear ion trap as the first analyzer and an Orbitrap as the second analyzer (LTQ-Orbitrap). Some MS/MS instruments have a single “ion trap” analyzer that traps ions and functions as the collision chamber. Ion traps are relatively inexpensive but they lack the sensitivity, versatility, and mass range of most tandem instruments. The Fourier transform mass spectrometer (FTMS), a very expensive but very powerful instrument, has extremely high mass accuracy. An alternative to CAD for fragment ion production has been developed for FTMS. This is called electron capture dissociation (ECD) and involves the capture of an electron by the multiply charged molecular ion. A related technique, called electron transfer dissociation (ETD) was subsequently developed for tandem instruments. Both ECD and ETD yield radical cations generated by different fragmentation pathways than ions produced by CAD. ECD and ETD are especially useful for defining glycosylation sites in glycopeptides.

Although underivatized glycans can be analyzed by MALDI- or ESI-MS/MS, superior data can be obtained if the glycans are derivatized before MS analysis. Derivatization methods can be broadly divided into two categories: (1) “tagging” of reducing ends and (2) protection of all functional groups. Commonly used tagging reagents include p-aminobenzoic acid ethyl ester, 2-AP and 2-AB. This type of derivatization facilitates chromatographic purification and enhances useful fragmentation. Protection of hydroxyl groups by permethylation is by far the most important type of full derivatization used in glycan MS (although with accompanying destruction of acetyl esters and other base-labile functional groups during the derivatization process). In MS/MS experiments, permethylated derivatives form abundant fragment ions produced by cleavage at susceptible glycosidic linkages, notably on the reducing side of each HexNAc residue. In N- and O-glycans, this preferred fragmentation unambiguously establishes the antennae sequences.

Broadly speaking, the unique strengths of MS can be exploited in two general ways in glycobiology. The first is to obtain detailed characterization of purified individual glycans, or mixtures of glycans. In this type of study, it is essential to acquire sufficient rigorous data to define structure unambiguously; many different MS-based experiments will be required, often complemented by NMR, linkage analysis, and profiling of enzyme digests. The second is to pursue glycomics investigations in situations in which it may not be essential to define structures fully, and high-throughput glycomic profiling or mass-mapping procedures can be exploited (Chapter 51).

NMR Spectroscopy

NMR spectroscopy is a powerful tool capable of full de novo structural characterization of isolated glycans and simple glycan mixtures. Among its several advantages are its broad applicability (glycans contain an abundance of protons, the most easily detected magnetic nucleus), its nondestructive nature, the quantitative relationship between resonance intensity and residue concentration, the diverse set of experiments available, and the ability to return information on both primary and tertiary structure. Its principal limitation is sensitivity, typically requiring several nanomoles of glycan for even the simplest experiment directed at structure determination. However, there are emerging experiments using methods, such as dynamic nuclear polarization, that provide several orders of magnitude improvement in sensitivity for applications that can monitor rapid in vivo conversions of metabolites such as glucose and fructose. Other limitations for both structural and metabolic applications are the relatively high cost of spectrometers and the level of expertise required for acquiring and interpreting NMR spectra.

When small amounts of sample are available and the source of the sample is known, the acquisition of a 1D proton NMR spectrum, combined with simple 2D coupling correlated spectroscopy (COSY) or total correlation spectroscopy (TOCSY), can frequently identify a glycan. Figure 50.3 shows a 1D proton NMR spectrum collected on 30 µg of PNGase F-released N-glycans from the Fc fragment of a human IgG1 antibody engineered to have a near homogeneous glycan compostion. The collection required ∼20 min on a currently state-of-the-art spectrometer (900 MHz with a cryoprobe). The anomeric proton resonances appear in a well-resolved region of the spectrum and show characteristic doublets with a splitting that is small for most α anomers and is significantly larger for most β anomers. Thus, a first glance at the ¹H-NMR spectrum gives a good estimate of how many residues there are (one anomeric proton per residue except for Sia), and how many of them belong to each anomeric type. Connections from the anomeric (H1) resonances to the H2 and subsequently other resonances can usually be determined from simple COSY or TOCSY data, which show chemical shifts (resonance positions) for these protons as cross peaks on columns or rows emanating from an anomeric resonance (an ∼24-h acquisition in the example given). The downfield (higher ppm) position and small couplings of the H2 resonances connected by cross peaks to three of the α-anomeric resonances identify these as belonging to core Man residues of an N-glycan. These, including other easily resolved resonances (historically called reporter group resonances), are sensitive to both monosaccharide type and linkage type, and often prove sufficient to identify a glycan, particularly when options are restricted by knowledge of biosynthetic pathways in the system under study.

FIGURE 50.3.

Section of a 900-MHz proton nuclear magnetic resonance (NMR) spectrum of N-glycans released from the Fc of human IgG1. (A) 1D proton NMR spectrum, ∼30 µg in ∼300 µL of D₂O. (B) COSY spectrum of the same sample showing connectivities (more...)

Accessible structures are more diverse in other organisms (e.g., bacteria and plants), and it is often advantageous to include ¹³C chemical shifts. These can be correlated with resonances of directly bonded protons using 2D heteronuclear single quantum coherence (HSQC) spectra. ¹³C is magnetically active, but ¹³C spectroscopy is less sensitive than proton spectroscopy, so chemical shifts are indirectly detected through protons. Even with the improved detection, ¹³C is just 1% naturally abundant, and unless the sample is isotopically enriched, approximately two orders of magnitude more material is required for collection of these data. Web-based tools that rely on empirical rules or databases that correlate chemical shifts with structural features now exist, and these allow facile prediction of possible structures from 1D and 2D NMR data sets (e.g., CASPER).

When unprecedented structures emerge, it is also possible to perform a de novo structure determination using NMR data alone. This usually begins with a full assignment of the ¹H and ¹³C resonances for each residue of a glycan using a combination of COSY, TOCSY, and HSQC spectra. In addition, the two-dimensional heteronuclear multiple-bond correlation (HMBC) experiment is used to detect through-bond coupling between the anomeric proton and the carbon atom on the opposite side of the glycosidic linkage. This is, however, an even less sensitive and more time-consuming experiment. It is possible to avoid both ¹³C-based HSQC and HMBC experiments by substituting a 2D nuclear Overhauser effect (NOE) spectrum or its rotating frame analog, an ROE spectrum. These experiments correlate proton resonances based on interproton distances, as opposed to through-bond couplings; they typically show cross peaks for pairs of protons that are within ∼4 Å of one another. Most connections are between protons in the same residue, but typically an extra connection will be between an anomeric proton and a proton in another residue. This proton is frequently across the glycosidic bond at the linkage site, providing a way to assemble residues into a complete structure. However, caution must be exercised. The identification of linkage sites is not as definitive as with ¹³C-based experiments, because the distance for the trans-glycosidic pair is dependent on glycosidic torsion angles, and these NOE cross peaks can be weak. Also, other non-linkage site protons can come within NOE distance and give false identification.