Glycobiology, 2000, Vol. 10, No. 7 669-689
© 2000 Oxford University Press
Exploring the glycan repertoire of genetically modified mice by isolation and profiling of the major glycan classes and nano-NMR analysis of glycan mixtures
Departments of Medicine and Cellular and Molecular Medicine and 3Glycobiology Research and Training Center, Howard Hughes Medical Institute, University of California San Diego, La Jolla, CA 92093-0687, USA
Received on October 27, 1999; revised on January 18, 2000; accepted on January 28, 2000.
 |
Abstract |
|---|
 Top Abstract IntroductionResults and discussionMaterials and methodsAcknowledgmentsAbbreviationsReferences |
|---|
The production of mice with genetic alterations in glycosyltransferases has highlighted the need to isolate and study complex mixtures of the major classes of oligosaccharides (glycans) from intact tissues. We have found that nano-NMR spectroscopy of whole mixtures of N- and O-glycans can complement HPLC profiling methods for elucidating structural details. Working toward obtaining such glycan mixtures from mouse tissues, we decided to develop an approach to isolate not only N- and O-glycans, but also to separate out glycosphingolipids, glycosaminoglycans and glycosylphosphatidylinositol anchors. We describe here a comprehensive Glycan Isolation Protocol that is based primarily upon the physicochemical characteristics of the molecules, and requires only commonly available reagents and equipment. Using radiolabeled internal tracers, we show that recovery of each major class of glycans is as good or better than with conventional approaches for isolating individual classes, and that cross-contamination is minimal. The recovered glycans are of sufficient purity to provide a "glycoprofile" of a cell type or tissue. We applied this approach to compare the N- and O-glycans from wild type mouse tissues with those from mice genetically deficient in glycosyltransferases. N- and O-glycan mixtures from organs of mice deficient in ST6Gal-I (CMP-Sia:Galß1–4GlcNAc
2–6 sialyltransferase) were studied by the nano-NMR spectroscopy approach, showing no detectable 2–6-linked sialic acids. Thus, ST6Gal-I is likely responsible for generating most or all of these residues in normal mice. Similar studies indicate that this linkage is very rare in ganglioside glycans, even in wild-type tissues. In mice deficient in GalNAcT-8 (UDP-GalNAc:polypeptide O-Ser/Thr GalNAc transferase 8), HPLC profiling indicates that O-glycans persist in the thymus in large amounts, without a major change in overall profile, suggesting that other enzymes can synthesize the GalNAc-O-Ser/Thr linkage in this tissue. These results demonstrate the applicability of nano-NMR spectroscopy to complex glycan mixtures, as well as the versatility of the Glycan Isolation Protocol, which makes possible the concurrent examination of multiple glycan classes from intact vertebrate tissues. Key words: glycan/mice/glycosyltransferase/NMR
|
Introduction |
|---|
|
TopAbstractIntroductionResults and discussionMaterials and methodsAcknowledgmentsAbbreviationsReferences |
|---|
Eukaryotic cells express several classes of oligosaccharides attached to proteins or lipids (Stults et al., 1989
; Ferguson, 1992; Furukawa and Kobata, 1992; Hart, 1992; Lis and Sharon, 1993; Margolis and Margolis, 1993; Hascall et al., 1994; Schnaar, 1994; Varki and Freeze, 1994). Animal glycans can be N-linked via ß-GlcNAc to Asn (N-glycans), O-linked via -GalNAc to Ser/Thr (O-glycans), or can connect the carboxyl end of a protein to a phosphatidylinositol unit (GPI-anchors) via a common core glycan structure (Ferguson, 1992; Furukawa and Kobata, 1992; Hart, 1992; Lis and Sharon, 1993; Varki and Freeze, 1994; Burda and Aebi, 1999). In heparan sulfate (HS), chondroitin sulfate (CS), and dermatan sulfate (DS) proteoglycans, the linkage region of the glycosaminoglycan (GAG) chain has a ß-Xyl residue O-linked to Ser (Hascall et al., 1994; Esko and Zhang, 1996). In contrast, keratan sulfate (KS) is a GAG chain extended from N-glycan or O-glycan cores (Margolis and Margolis, 1993; Hascall et al., 1994; Iozzo, 1998). Nuclear and cytoplasmic glycosylation also exists, with the most common being ß-GlcNAc-O-Ser/Thr (O-GlcNAc) (Holt et al., 1987; Hart et al., 1989). Glycosphingolipids (GSLs, often called glycolipids) have a glycan linked to a ceramide moiety via ß-Glc (Glc-Cer) or ß-Gal (Gal-Cer) (Hakomori, 1990; Zeller and Marchase, 1992; Sandhoff and Klein, 1994; Karlsson, 1995). Free glycan chains also occur, including sialyloligosaccharides (Kunz and Rudloff, 1993), hyaluronan (HA) (Laurent and Fraser, 1992), and degradation fragments of GAG chains (Bai et al., 1997).
N-Glycans are grouped into high mannose, complex, and hybrid types (Furukawa and Kobata, 1992; Varki and Freeze, 1994; Stanley and Ioffe, 1995), with ß-GlcNAc branches in complex N-glycans giving di- or multi-antennary units (Schachter, 1991). Vertebrate O-glycans include seven or more different core structures (Schachter and Brockhausen, 1992), and are often found in clusters on polypeptides called mucins (Verma and Davidson, 1994; Devine and McKenzie, 1992; Rose, 1992; Strous and Dekker, 1992). Cells may also contain lipid-linked oligosaccharide precursors from the N-glycan and GPI-anchor biosynthesis pathways (Ferguson, 1992; Furukawa and Kobata, 1992; Burda and Aebi, 1999). Neutral and acidic GSLs can have a variety of extensions arising from the Glc-Cer or Gal-Cer core regions (Hakomori, 1990; Zeller and Marchase, 1992; Sandhoff and Klein, 1994; Karlsson, 1995). While core regions are unique for each class, elongation can involve similar outer chains. For example, N- and O-glycans and GSLs often carry subterminal Galß1–4GlcNAcß1- (so-called LacNAc) units, which can sometimes be repeated (polylactosamines), or less commonly, Galß1–3GlcNAcß1- or GalNAcß1–4GlcNAcß1- (so-called LacdiNAc) units (Smith et al., 1992; Baenziger, 1996). Some of these units are capped by sialic acid (Sia), Fuc, -Gal, or ß-GlcA (Chan et al., 1991; Jungalwala, 1994; Varki and Freeze, 1994). Further diversity arises from modifications like sulfation (Smith et al., 1992; Hooper et al., 1996) and variations of Sia structure (Troy, 1992; Varki, 1992; Rutishauser, 1996; Kelm and Schauer, 1997). GPI-anchors can also carry various side chain substituents, including phosphoethanolamine units, Man, GalNAc, or Gal residues, or even sialylated LacNAc units (Ferguson, 1992). The outer regions of proteoglycans (Margolis and Margolis, 1993; Hascall et al., 1994; Salmivirta et al., 1996) are unbranched GAG chains, which are polymers of a disaccharide with a hexosamine and a uronic acid (GlcA or IdoA), or in the case of KS, Gal. With the exception of HA, such GAG disaccharides can be O-sulfated at various positions. In HS, some GlcNAc residues are N-de-acetylated or N-sulfated (Salmivirta et al., 1996).
Since glycans are involved in interactions with the environment, other cells or bioeffectors (Rademacher et al., 1988; Ferguson, 1992; Varki, 1993; Hakomori, 1990; Schnaar, 1994; Stanley and Ioffe, 1995; Baenziger, 1996; Crocker and Feizi, 1996; Gahmberg and Tolvanen, 1996; Rutishauser, 1996; Dennis et al., 1999; Reuter and Gabius, 1999), their isolation and structural characterization are of great interest. However, there is limited information about the glycan repertoire of whole cells, tissues, and physiological fluids. Indeed, most reported structures of vertebrate N- and O-linked glycans have been from readily accessible sources such as mucins, plasma glycoproteins or erythrocytes. Recently, intact mammalian tissues such as brain and lung have been submitted to total N-glycan release, aiming to generate "tissue oligosaccharide libraries" (Wing et al., 1992; Norgard-Sumnicht et al., 1995; Fujimoto et al., 1999). In two instances (Wing et al., 1992; Norgard-Sumnicht et al., 1995), such libraries included many unexplained anionic molecules. Thus, the oligosaccharide repertoire of mammalian organs may be even more extensive than summarized above. Investigators who study intact tissues typically focus on a specific class of glycans, and methods for their isolation and purification usually discard the other classes (Kobata and Furukawa, 1992; Hascall et al., 1994; Rudd and Dwek, 1997; Geyer and Geyer, 1998). In some situations, it would be desirable to isolate all glycan chains of all major groups of glycoconjugates in parallel. Examples are mice (Marth, 1994) or cell lines (Esko, 1991; Stanley, 1992) with genetically engineered defects in glycosylation. To compare such mutants to the wild-type situation we have developed a "Glycan Isolation Protocol" for the isolation and fractionation of all major classes of glycans from vertebrate cells or tissues, and a novel nano-NMR approach to analyzing complex glycan mixtures. These approaches have been applied here to analyze tissues from two types of mice with genetically induced changes in glycosylation.
|
Results and discussion |
|---|
|
TopAbstractIntroductionResults and discussionMaterials and methodsAcknowledgmentsAbbreviationsReferences |
|---|
Section 1. Developing a nano-NMR approach to analyzing glycan mixtures
NMR analysis of whole glycan mixtures from fetuin by 1H-NMR spectroscopy.
1H-NMR spectroscopy is regarded as a powerful technique for the structural characterization of glycopeptides or free glycans obtained from glycoconjugates (Vliegenthart et al., 1983
; Van Halbeek, 1993, 1994). However, complete structure determination by 1H-NMR requires a glycan’s availability in virtually pure state. Thus, the traditional approach has been to separate out each component for individual study by NMR spectroscopy. Despite its successes, this approach is limited in several ways: (1) separating closely related carbohydrate structures from each other, to the level of >95% purity, is a laborious effort at best, and simply impossible in many cases; (2) minor components in the mixture may be lost during the separation process, or cannot be recovered in amounts sufficient for NMR analysis; and (3) the quantitative relationship between various components of a mixture may get compromised during separation, making it difficult to obtain a global overview of the structural data, including the ratios of components in the starting mixture. Thus, we felt that there is enough to be gained from analyzing mixtures of glycans by 1H-NMR to offset the decrease in detailed structural information that can be obtained from such complex spectra.
Mixtures of 2 to 3 glycans have been previously examined by NMR for full structural characterization (see, for example, Spellman et al., 1989, 1991; Watson et al., 1994), but only when complete separation had not been possible. The larger the number of component glycans in a mixture, the less information can be gained about each individual component. On the other hand, global structural information about the presence of certain epitopes and their relative ratios in a mixture of glycans should be attainable from the spectra of mixtures. For example, we reasoned that it should be possible to obtain partial structural information from 1H-NMR spectra of whole complex mixtures of N- and O-glycans, particularly in cases where the purpose is to compare two closely related glycoprotein/glycopeptide samples. To deal with the issue of limited amounts of glycan samples, we used nano-1H-NMR spectroscopy (at 500 MHz) which we have shown previously (Manzi et al., 1995) to be capable of improving sensitivity up to 10-fold over conventional, 5 mm probe NMR at the same magnetic field strength. To test the concept, we initially studied bovine fetuin, a well-known glycoprotein with at least 30 structurally unique N- and O-glycan chains that have been extensively studied by many laboratories using diverse analytical methods (Baenziger and Fiete, 1979; Nilsson et al., 1979; Takasaki and Kobata, 1986; Green et al., 1988; Hardy and Townsend, 1988; Cumming et al., 1989; Hermentin et al., 1994; Ishii-Karakasa et al., 1997; Kotani and Takasaki, 1997; Chen et al., 1998; Mechref and Novotny, 1998). We obtained whole mixtures of N- and O-glycans from fetuin (Sigma) via automated hydrazinolysis in the N- and O-mode, followed by re-N-acetylation and clean-up. Starting from 2 mg of fetuin, the typical yield was ~100 µg of mixed N- and O-glycans. Figure 1 shows the spectrum obtained when this glycan mixture was analyzed by nano-probe 1H-NMR spectroscopy at 500 MHz. The presence and relative abundance of certain structural elements could be assigned from the structural reporter group signals that remained individually observable, despite the complexity of the mixture. The following structural features could be confidently assigned: (1) the overall ratio of Sia2–3Gal versus Sia2–6Gal structural elements was determined to be 2:1, by integration of the triplets corresponding to H3ax of 2–3 and 2–6 Sia, at 
1.80 and 1.72, respectively, and independently of the H3eq signals at 2.76 and 2.67, respectively; (2) the ratio of di-antennary versus tri-antennary N-glycans was estimated to be 1:3, based on the relative intensities of the Man H2 signals at 4.25 and 4.21, respectively; (3) Fuc, in either 1–6 or 1–3 linkage to GlcNAc residues or 1–2 linkage to Gal residues was undetectable, by virtue of the absence of Fuc CH3 signals in the 1.17<<1.28 region; (4) Neu5Gc was not present, as judged from the absence of the typical, sharp CH2 singlet at 4.11–4.12 (see below); and finally, (5) the apparent ratio of N- to O-glycans was found to be 3:1, based on integration of the H1 doublets of the -anomers of the reducing-end GlcNAc and GalNAc residues at 5.18 and 5.21, respectively. (To reconcile this with the theoretical ratio of 1:1, it is not sufficient to just take into account that the ratios of :ß anomers for N- and O-glycans are 67:33 and 55:45, respectively (Herlant-Peers et al., 1981; Spellman et al., 1991); it is likely that the relative yield of intact O-glycans in the automated hydrazinolysis procedure was lower than that of the N-glycans). On the other hand, the spectrum did not give unambiguous evidence regarding the occurrence of the Neu5Ac2–6(Galß1–3)GlcNAc structural element, which is known to be present in a small proportion of the triantennary N-glycans (Green et al., 1988; Cumming et al., 1989); nor was it possible to distinguish between Sia2–3Galß1–3 and Sia2–3Galß1–4 elements, or between Sia2–6Gal and Sia2–6GalNAc units. The NMR parameters for each of the aforementioned elements are known to be unique (Vliegenthart et al., 1982; Bernard et al., 1983; Lecat et al., 1984; Green et al., 1988), yet their differences are apparently too subtle to be observable in the spectrum of this mixture. Nonetheless, the data indicate that, at the level of glycan complexity present in a glycoprotein like fetuin, it is possible to study a whole mixture of N- and O-glycans by 1H-NMR spectroscopy and obtain substantial structural information.
|
Section 2. Development of a glycan isolation protocol for tissue and cells
To apply this novel NMR approach towards comparing glycosylation in normal and mutant mouse tissues, it is necessary to obtain mixtures of N- and O-glycans in a relatively pure state (i.e., free of other glycan types and of other 1H-containing cellular molecules). Also, mutations in glycosyltransferases could cause expected or unexpected changes in other classes of glycans, e.g., in GSLs. Thus, in developing a generally applicable approach, we decided to attempt the concurrent extraction, fractionation and isolation of all of the major classes of glycans from cells and tissues. The envisaged approach would allow the characterization of any cell type or tissue by the relative proportion of structural species within each of the major classes of glycans that it harbors, i.e., it would provide a "glycoprofile" of the cell or tissue. Alternatively, if the immediate aim is to isolate a particular class of glycans, the approach should leave the other classes in a form that could be stored and examined at a later time, as needed. Another goal was to design an approach that, with minimal modification, would be generally applicable to a wide range of biological samples (i.e., cells, organs, body fluids) where combinations of different glycoconjugates are present. The ideal protocol should satisfy the following criteria: maximum possible recovery of each group of glycans, without any selective losses of particular subsets present in the starting material; minimal contamination of one group of glycoconjugates with another; assure stability of the most labile sugar residues (e.g., Sias), and the most labile modifications (i.e. sulfate esters, phosphate esters, O-acetyl esters); and individual fractions should be obtained in a form that does not impair subsequent glycan release and/or profiling.
The Glycan Isolation Protocol described here comes close to satisfying these criteria. The overall scheme is outlined in Figure 2, and described in detail under Materials and methods. It involves the initial extraction of GSLs and other glycolipids (along with phospholipids and sterols), SDS-mediated solubilization and subsequent proteolysis of denatured glycoproteins and proteoglycans, recovery of GPI-anchors as lipid-linked glycopeptides, and finally, the separation of N- and O-glycan containing glycopeptides from high-charge GAG-containing glycopeptides. Glycans from each fraction can subsequently be released and profiled, and selected components of each can be subjected to detailed analysis. Along the way, fractions containing low-molecular-weight molecules like monosaccharides, sugar nucleotides, free oligosaccharides, and small glycopeptides can be saved separately. Certain variations and optional extensions of the protocol are feasible (see Figure 3). As described below, the approach was validated using a set of radiolabeled standards representing each of the major classes of glycans to be separated from each other. First, we will briefly outline the rationale for the major decisions taken in setting up the approach.
|
|
Determining the extraction sequence and other conditions.
Using normal mouse thymus, we initially tested two alternatives: extracting lipids first and then solubilizing the proteins for glycopeptide preparation; or preparing glycopeptides first and later partitioning the lipids into an organic solvent. With the second approach, ~40% of glycopeptides were extracted with the lipids. Lipid extraction was therefore established as the first step. An added advantage is that all biologically active proteins (e.g., glycosidases) from a tissue would be inactivated by organic solvent denaturation at the outset. Also, GPI-anchors attached to proteins would not be extracted, allowing separation at a later stage, after proteolysis. A minor disadvantage is the possibility that some highly charged GSLs might not be extracted in the initial organic phase. However, these would be recovered after proteolysis in a second acidified organic extract, along with GPI anchors. During the initial phases of developing the approach, the recovery of radiolabeled standards representing each of the major classes of glycans was studied at various steps, either when used as pure populations or when spiked into a typical tissue (kidney or brain) prior to extraction. This allowed rational decisions to be made regarding the modification of various steps, to maximize recovery and separation. By maintaining pH 7.4 and room temperature (RT) throughout the protocol we also prevented losses of labile groups like O-acetyl esters (Varki and Diaz, 1984
).
Separating protein- or lipid-bound glycans from soluble glycans.
Cells and body fluids contain pools of monosaccharides, sugar phosphates and sugar nucleotides and occasionally, free glycans. Such small molecules are best removed or separated from protein-bound glycans prior to protease treatment, either by dialysis using a large molecular weight cut-off (MWCO) membrane, or by membrane ultrafiltration. We found no difference between conventional dialysis and ultrafiltration on membrane cartridges for elimination of small molecules, or for the subsequent separation of small glycopeptides from the larger species (see below). Ultrafiltration is advantageous as the initial step when analyzing biological fluids that contain free oligosaccharides (e.g., milk) allowing them to be recovered in a reasonable volume for further fractionation and analysis. Likewise, during isolation of small O-type glycopeptides after the proteolysis step, it is convenient to have them recovered in the small volume obtained by ultrafiltration. If recovery of either of these low-molecular-weight fractions is not of interest, conventional dialysis is simpler, albeit with greater risk of accidental sample loss.
Approaches for glycopeptide production. To obtain reliable separations based on intrinsic properties of the glycans, the size of attached peptide needs to be minimized. Initial experiments using mouse brain or kidney as carriers for radiolabeled standards indicated similar effectiveness of Proteinase K and Pronase, as long as SDS was used to ensure complete solubilization. Recombinant Proteinase K was adopted since, unlike Pronase, it has no contaminating enzymatic activities. In the absence of SDS, complete solubilization of the proteins denatured by organic extraction was difficult. Also, while recovery and distribution of N- and O-glycans from a standard glycoprotein ([3H-sialyl]fetuin) was good even without SDS, radiolabeled GPI-anchor glycopeptides and mucin-type glycopeptides were recovered in the wrong fractions (data not shown). The poor recovery of GPI-anchor glycopeptides in the acidic organic extract is probably because attached proteins could not be completely digested without SDS denaturation.
Essentially complete solubilization of the delipidated samples was observed with 1% SDS for all cells and tissues studied. Furthermore, SDS could be effectively removed as a potassium salt precipitate after proteolysis. We did explore alternatives to SDS such as reduction-alkylation with dithiothreitol followed by iodoacetamide. While subsequent GPI extraction after proteolysis occurred at acceptable levels, the recovery of the 35S-labeled GAG-type glycopeptides was surprisingly reduced, evidently due to a ~20% loss of sulfate esters. Although this was not observed with dithiothreitol alone, it was not practical to maintain a reducing agent throughout the protocol (including the DEAE column elution). Also, random exposure to an oxidizing environment could allow artifactual disulfide bond formation between free thiols. Finally, the reducing agent could not be easily eliminated from the fraction containing the small O-type glycopeptides. For all these reasons, reduction alone or reduction-alkylation were not viable alternatives or adjuncts to denaturation of nonreduced proteins by SDS. Of course, reduction-alkylation could always be done on the isolated DEAE glycopeptide fractions of interest at the end, if this were necessary to optimize certain separations.
Recovery and distribution of sialic acids and other monosaccharides in different fractions obtained using the optimized protocol.
To demonstrate the feasibility and efficacy of the approach, we studied two types of tissues from the mouse: brain (a "soft" tissue, very rich in lipids) and kidney (a "hard" tissue, enriched in epithelial cells and extracellular matrix). The optimized protocol was applied to such samples, and aliquots obtained at each step (see Figure 2) were monitored for total Sia content (by the DMB-HPLC assay after saponification of O-acetyl esters and mild acid hydrolysis). Aliquots of the final fractions were also subjected to strong acid hydrolysis and studied for monosaccharide content by HPAEC-PAD (this was not feasible with some of the less pure intermediate fractions, because of interfering impurities). Representative results of these analyses are shown in Tables I and II. The distribution of Sias, neutral monosaccharides and hexosamines was generally appropriate for the predicted composition of the individual fractions. Because of the sensitivity of the DMB-HPLC assay and the lack of interference by other molecules, routine monitoring of Sia recovery was most useful to ascertain protocol reproducibility at different steps. Overall glycan recoveries from the starting homogenate ranged from 45–85%, with improved recoveries typically seen after an operator had gained experience with the protocol.
|
Recovery and distribution of standard glycoconjugates submitted to the optimized protocol. Bovine fetuin (which contains various kinds of sialylated O-glycans and N-glycans) was radiolabeled in the Sia side chains via mild periodate/NaB[3H]4 reduction, and then either subjected directly to automated hydrazinolysis for release of glycans, or mixed with unlabeled brain tissue homogenate and put through the optimized protocol, followed by automated hydrazinolysis of the glycopeptide fractions. As expected, essentially all of the glycans were recovered either in the second ultrafiltration (SLMW3, the small O-glycans, containing ~1/3 of the recovered radioactivity), or in the fraction eluted with 100 mM NaCl from DEAE (GP0–100, comprising the remainder of the O-glycans and all the N-glycans, containing the remaining ~2/3 of the recovered radioactivity). The combined HPAEC profile of recovered glycans was similar to that of the starting material, indicating no major selective loss or structural change of any of the N- or O-glycans during the process (data not shown).
To directly monitor the recovery of individual classes of glycans in the presence of total tissue materials, 10,000 c.p.m. of radiolabeled standards representing each major class of glycan (see Materials and methods) were added to separate aliquots of the minced tissues (40–70 mg wet weight brain/aliquot, 35–60 mg wet weight kidney/aliquot), and the mixtures were submitted to the optimized protocol. As shown in Table III, the distribution of radioactivity indicates a reasonable recovery of the standards in the expected fractions, with relatively little cross-contamination (some differences in relative distribution do occur, depending upon whether the "carrier" tissue is brain or kidney). An exception was that a portion of the mucins were recovered in the GP100–1000 fraction, instead of the GP0–100 fraction. This is presumably because the clustered O-linked glycans on the mucins have a charge density similar to that of the GAGs (this problem can be partly alleviated by adding a 300 mM NaCl elution step prior to the 1 M NaCl step, see Figure 2, and below). The lowest yield was seen for the GPI anchors, where about ~50% of the starting material was found in the SLE fraction. However, this is not much worse than the yield obtained using the conventional protocol (extraction with Triton X-114) for isolating these glycans (~70% in our hands, data not shown). We also noted that the standard 3H-labeled GAGs derived from CHO cells (namely, CS and HS glycopeptides) underwent some degradation during the protocol. As shown in Table III, ~20% of the radioactivity was found in the PLMW10 fraction, instead of in the GP100–1000 fraction as expected. This is not an artifact due to the quality of the standard or to fragmentation occurring during the homogenization (based on control data with pure standards, not shown). Thus, significant degradation had occurred during the short period of time after mixing the standard into the minced tissue, and prior to adding organic solvent (typically 5–10 min), presumably due to endogenous GAG-degrading enzymes. We have not pursued this matter further, but it indicates that endogenous glycan-degrading enzymes must be considered, especially at the early stages of the protocol.
|
Overall, these data indicate that with minor imperfections, the optimized Glycan Isolation Protocol yields acceptable recovery and level of purity when known standards are added to tissues prior to fractionation. It is reasonable to expect that the endogenous glycans will behave in a similar fashion to the added-back standards. Further evidence that this is indeed the case is provided in the companion paper (Norgard-Sumnicht et al., 2000
), where metabolically radiolabeled endogenous glycans from cultured cells are recovered using the same protocol. Studies were also performed using other glycoconjugates such as HA and KS, monitoring recoveries through the individual steps with appropriate assays (data not shown). The overall conclusion from these studies is that the Glycan Isolation Protocol as described under Materials and methods is capable of achieving most of the original objectives: near-complete separation of the various classes of glycans, with a recovery of each of them comparable to conventional methods optimized for the isolation of individual classes. However, some critical comments are in order. First, small free oligosaccharides and sugar nucleotides partition partially into the PLE fraction. If desired, dialysis or ultrafiltration can be used to recover these molecules from this fraction. Second, glycopeptides with clustered O-linked chains (mucin-type glycans), which are protected from complete proteolysis by the high density of glycosylation, do not separate completely from the GAG chains. However, by eluting the DEAE-column in a stepwise fashion with 300 mM NaCl after the 100 mM NaCl step, it is possible to achieve substantial separation of these molecules from HS and CS chains (depending upon the degree of sulfation of the latter chains). Alternatively, ß-elimination can release both types of chains before separation on the basis of size and/or charge. Finally, other molecules with extensive size/charge heterogeneity (i.e., HA, KS) are distributed amongst the DEAE fractions. However, this is actually an advantage, since such small GAG chains would be missed when using conventional strategies aimed at isolating "typical" molecules. Contamination of the various fractions by other cellular constituents. Various low-molecular-weight components of cells (salts, nucleotides, amino acids, etc.) are recovered during the initial ultrafiltration/dialysis step and/or in the PLE fraction along with the monosaccharides, sugar nucleotides, and small free oligosaccharides. Very hydrophobic proteins (so-called "proteolipids") might be found in the PLE fraction. Except for regions carrying clustered O-linked chains, all remaining water-soluble polypeptides are completely digested by Proteinase K after boiling in SDS. Of course, small portions of peptide remain attached to the GPI-anchors and to the N- and O-glycans and to GAG chains. Most of the cellular phospholipids and neutral lipids are found in the initial organic extract, and the remainder are recovered in the second acidic organic extraction (as detected by HPTLC, data not shown). Fortuitously, all of the nucleic acids were recovered in the SLE (secondary lipid extract; see Figure 4), presumably because the acidification of the extraction mixture protonates the phosphodiester bonds in DNA and RNA. Regardless of the reason, the extraction of nucleic acids at this step assures that they do not contaminate the GAG chains, which are recovered later with a high salt elution from the DEAE column.
|
Options
, variations and extensions of the approach. Depending on the question at hand, several options and variations of the Glycan Isolation Protocol can be used to achieve specific purposes (see Figure 3 and Materials and methods).
Profiling of released oligosaccharides. Various profiling methods were used to monitor the success of the protocol (see Materials and methods). In the remainder of this paper, we will focus mainly on profiling the N- and O-glycans of mouse tissues. Optimization and automation of hydrazinolysis has allowed uniform and simultaneous release of both N- and O-linked glycans in acceptable yield (Patel et al., 1993). HPAEC is a suitable profiling strategy for such free N- and O-glycans derived from the low-charge (GP0-100) glycopeptide fraction, since even isomeric forms differing in only one linkage can be resolved from each other (Hardy and Townsend, 1994). Indeed, as shown in Figure 5, the purity of the N- and O-glycans released from this fraction by automated hydrazinolysis is sufficient to obtain a clean HPAEC-PAD profile that is relatively reproducible from preparation to preparation. Also, different tissues give markedly different profiles of released oligosaccharides.
|
Some variations in the profiles between individual samples from the same tissue do occur (see Figures 5 and 6 for two separate examples). This is most likely explained by individual biological variation, e.g., recent data indicate that expression patterns of sialyl linkages in the rat brain change with aging (Sato and Endo, 1999
). Another possibility is the relative amount of the hydroxylated Sia, N-glycolyl-neuraminic acid (Neu5Gc). Molecules containing this residue instead of N-acetyl-neuraminic acid (Neu5Ac) show markedly higher retention times on HPAEC-PAD. Thus, relatively small changes in the percentage of Neu5Gc could significantly change the profile. In this regard, we have noted that not all the Neu5Gc present in the original sample is transformed into Neu5Ac during the hydrazinolysis/re-N-acetylation procedure (data not shown). Thus, a complementary approach is the enzymatic release of the Sias followed by profiling of the underlying neutral oligosaccharides by HPAEC-PAD using a gradient designed to separate such chains.
Section 3. Application of the glycan isolation protocol to transgenic mice
Analysis of glycans from mice with a null mutation in the ST6Gal-I sialyltransferase. Following disruption of a gene encoding a specific glycosyltransferase in intact mice, the question arises if the predicted change in glycosylation does indeed occur in the genetically altered animal. The failure to find the predicted change in glycosylation in a particular tissue would indicate the presence of other enzymes capable of producing the same structure, while a complete loss of the pertinent structural element would prove that the gene deleted is the only one capable of biosynthesizing it in the tissue/cell type being examined. For example, the genetic disruption of the then only known ß-galactosyltransferase (Asano et al., 1997; Li et al., 1997) did not result in elimination of all ß-Gal residues (Kotani et al., 1999), because of the existence of several additional isoforms of this enzyme (Lo et al., 1998). Furthermore, since most major glycoconjugate classes are synthesized in the Golgi-ER pathway and compete for the same sugar nucleotide pools, the question arises if unexpected changes occur in other classes of glycans. The protocol described above is presented as a first step towards addressing these issues.
ST6Gal-I (CMP-Sia:Galß1–4GlcNAc 2–6 sialyltransferase) is the only mammalian enzyme currently known to generate the Sia2–6Gal linkage and its minimally required acceptor is Galß1–4GlcNAcß (Tsuji, 1996). Despite the widespread constitutive or regulated expression of ST6Gal-I in various cell types (O'Hanlon et al., 1989; Wang et al., 1989; Jamieson et al., 1993; Lo and Lau, 1996) the phenotypic consequences of ST6Gal-I disruption in the mouse appeared to be limited to B cells, which also appeared to lose expression of the Sia2–6Gal glycan structural element as determined by loss of binding by Sia specific lectins (Hennet et al., 1998). It is conceptually possible that there are other ST6Gal enzymes in other tissues and cell types, transferases that could partly or fully compensate for the lack of ST6Gal-I expression. To ascertain the occurrence of glycans terminating in Sia2–6Gal in tissues other than B-cells, we studied the GP0–100 and GP100–300 fractions derived from the liver, brain, and kidney of normal and ST6Gal-I null (homozygote and heterozygote) mice. Since these fractions contain many structurally distinct glycans, they were not amenable to fractionation and purification of each individual component. However, as shown in Figure 7, 1H-NMR analysis of the glycopeptide mixtures present in the GP0–100 fractions of normal and mutant mouse liver indicated that the ST6Gal-I deficient mouse tissue lacks any detectable 2–6 linked Sia, whereas 2–3 and 2–6 linked Sia in the corresponding glycans from control tissues occur in the ratio of approximately 1:1. Not only do these fractions represent mixtures of many glycans (more than 50, as seen by HPAEC-PAD of these fractions after hydrazinolysis), they contain the glycans in peptide-bound form. Despite the added complication posed by the presence of amino acid signals in the 1H-NMR spectra, the relatively narrow line width and the characteristic patterns and well-defined positions of Sia H3ax and H3eq signals in the spectra enabled us to establish the presence of Sia2–3Gal and Sia2–6Gal units. Furthermore, it was possible to come to this conclusion with only ~200 g of the glycopeptide mixtures. Indeed, removal of the peptide portions by automated hydrazinolysis and sample clean-up gave glycan mixtures that did not give better quality spectra by 1H-NMR spectroscopy. Furthermore, the loss of glycan material during hydrazinolysis was ~50%, and the removal of 1H containing contaminants was apparently incomplete. The NMR spectra of the liver GP100–300 fractions (not shown) revealed the same trend as those of the GP0–100 fractions: no Sia2–6 was detectable in the material from ST6Gal-I null mice, whereas the spectrum of the wild-type material showed a minor proportion (~5%) in 2–6 linkage. The 500 MHz 1H-NMR spectra of the GP0–100 fractions from kidney and brain tissue (not shown) gave the same result as with liver. Taken together, these data indicate that ST6Gal-I is responsible for generating most or all of the Sia2–6Gal residues on N- and O-glycans in these tissues of normal mice (the lower limit of sensitivity for detection of 2–6 linked Sia is 2–3% compared to the level of 2–3 linked Sia). Thus, direct 1H-NMR analysis of complex mixtures of glycans can be an adequate tool to assess changes in the levels of specific Sia units in sialyltransferase-null mouse tissues, and to measure the relative ratios of sialic acids in different types of linkages.
|
The ST6Gal-I product is not detectable in normal mouse liver gangliosides. While the ST6Gal-I acceptor Galß1–4GlcNAcß1- is not found in mammalian brain GSLs, it is abundant in extraneural gangliosides (Stults et al., 1989
). However, the Sia2–6Galß1–4GlcNAcß1- product seems to be generally rare in the GSLs of normal human tissues, and is reported primarily in fetal and malignant samples (Taki et al., 1992a,b). Thus, it has been suggested that there may be a different enzyme responsible for generating this structure on GSLs (Taki et al., 1992a). To address whether there was a change in relative abundance of Sia linkage types in GSL glycans between ST6Gal-I control and null mice, we enriched the PLE fractions from brain and liver in larger-chain GSLs by Folch partitioning. This step is justified since the minimal length of a lacto-type ganglioside structure potentially affected by the mutation is a pentaglycosyl ceramide, namely, Sia2–6Galß1–4GlcNAcß1–3Galß1–4Glcß1–1Cer; this component and all larger/branched lacto-type gangliosides end up in the PLE Folch upper phase. After release by endoglycoceramidase, the mixtures of larger-size GSL glycans were subjected to 1H-NMR spectroscopy. The spectra shown in Figure 8 indicate that 2–6 linked sialic acids are hard to detect (i.e., they represent less than 2–3% of the 2–3 and 2–8 linked sialic acids combined) in the mixtures of GSL-derived oligosaccharides obtained from brain and liver PLE of control mice. While the absolute amounts of ganglioside oligosaccharides obtained per 100 mg of wet tissue from mouse brain far outweighed those from liver, the amount of GSL glycans obtained from 2 g of mouse liver was well within detection limits of nano-1H-NMR spectroscopy. The apparent absence of Sia2–6Gal containing gangliosides from liver in wild-type mice is remarkable, considering that ST6Gal-I expression in liver is very prominent, and glycoprotein-bound 2–6 linked sialic acids are quite common in liver (compare Figure 7). Thus, 2–6 linked sialic acids are very rare in the total GSLs of normal mouse brain and liver. We are currently searching for a normal mouse tissue wherein they are even detectable. Once we have found one or more examples of such a tissue, we can address whether the corresponding tissue(s) in ST6Gal-I null mice are deficient in ganglioside Sia2–6Gal. This should allow us to establish if the same ST6Gal-I enzyme that adds 2–6 linked sialic acids to N-and O-glycans on proteins also synthesizes GSLs with this linkage.
|
Mice with a null mutation in the GalNAcT-8 transferase do not show a major loss of O-glycans. In the case of mice with deficiency of GalNAcT-8 (UDP-GalNAc:polypeptide O-Ser/Thr GalNAc transferase 8) (Hennet et al., 1995
), we analyzed thymocytes, as well as brain and kidney samples, for the profile of N- and O-linked chains derived from the total GPU fraction. An example of the data obtained from the thymocytes is shown in Figure 6. In every case, we found that the region of the HPAEC-PAD profile expected to contain the O-glycans did not show any obvious change (although in some samples there appeared to be a decrease in the area of some of the O-glycan peaks). These data are best explained by the recent findings that there are several vertebrate polypeptide GalNAc-T enzymes capable of forming the GalNAcO-Ser/Thr linkage (Bennett et al., 1998; Hanisch et al., 1999). These isozymes show some degree of tissue-specific expression. Thus, it is possible that if all the tissues of the mouse were studied in the aforementioned manner, one might find a major change in the amount of O-glycans in some location. This possibility has not yet been pursued. Note that we also did not examine the fraction SLMW3, that can contain some of the small, nonclustered O-glycan chains.
|
Conclusions and perspectives
We have demonstrated the value of applying NMR spectroscopy to whole mixtures of oligosaccharides or glycopeptides from mouse tissues. When coupled with the improved sensitivity of the nano-NMR probe, this approach allows the study of complex analytical situations such as those presented by glycosyltransferase-deficient mice. The advantages and limitations of this approach need further exploration. The application of this NMR technique to mouse tissues required the development of a novel Glycan Isolation Protocol. Most published protocols for extracting glycoconjugates from tissues and cells focus on recovering one specific class of molecules. There have been previous attempts to simultaneously extract more than one class, but none have taken the comprehensive approach presented here. Even if the initial goal is only to study one or two of the classes of glycans, the present approach leaves other purified fractions, in case a new question subsequently arises. We have also deliberately based our approach exclusively on the physicochemical properties of glycans, and avoided the use of tools such as lectins and enzymes. While the latter can show exquisite specificity, they can also give misleading results, e.g., some N-linked chains are resistant to PNGaseF, and some lectins will not bind their target structures when the latter are modified by anionic groups like sulfate. Using a primary organic extraction followed by solubilization of precipitated molecules in SDS, we also avoided most artifacts caused by degradative enzymes that might be released upon tissue homogenization (a small amount of GAG degradation was seen despite this). Another advantage of the protocol is that it uses relatively inexpensive materials and tools available to most laboratories doing conventional molecular biology or biochemistry. Furthermore, we have found the approach to be relatively reproducible when performed by four different operators following a standardized written set of instructions. Notably, although the method has not been optimized for this purpose, we were able to selectively extract the nucleic acids in one particular fraction (away from the GAGs), with the DNA being of relatively high molecular weight.
The approach has some minor disadvantages and limitations. First, the whole process takes about one week, and involves many steps. However, up to eight samples can be processed in parallel, and there are large time gaps between steps, allowing the operator to pursue other activities in parallel. Second, the glycoproteins and proteoglycans are recovered as an SDS extract, and are therefore not suitable for isoelectric focusing or functional studies. However, prior to proteolysis, the extract is suitable for analysis by SDS-PAGE, and for antibody or lectin blotting studies. Third, some highly acidic GSLs may be recovered in the fraction meant to contain only GPI-anchor glycopeptides, and there is the potential for methyl esterification of sialic acids under these acidic conditions, in the presence of methanol. However, the fraction of such molecules is relatively small in most tissues, and other methods can be used to differentiate them from the GPI-anchors. Fourth, the use of exhaustive proteolysis limits the subsequent use of some N-glycanases which require some length of amino acids adjacent to the linkage region. Finally, the DEAE column gives incomplete separation of mucin glycopeptides from some GAG glycopeptides. However, as discussed earlier, a variety of nondegradative and degradative techniques can be used to achieve the desired separations.
Our immediate purpose in developing this approach was to study the outcome of genetic modulation of glycosyltransferases in mice. Given the rapid progress in identifying the genes encoding the glycosyltransferases, some might argue that it would be simpler to examine the expression pattern of their messenger RNAs as a means towards profiling glycosylation potential. However, there is now increasing evidence that expression of mRNA does not always correlate with the levels of a functional protein present in a given cell type (Page et al., 1999). In the case of glycosylation, even a direct measurement of glycosyltransferase activity is not sufficient. Unlike the synthesis of RNA and proteins, glycosylation is not exclusively template driven, and one cannot accurately predict the glycosylation phenotype of a given tissue or cell type from patterns of glycosyltransferase activity expression. For example, even though GlcNAcT-IV activity is barely detectable in rat liver, many of the major serum glycoproteins produced by the rat hepatocyte have multiantennary N-linked chains that include the ß1–4 branch synthesized by this enzyme (Schachter, 1991). Likewise, the ratio of 2–3 to 2–6 linked sialic acids was only moderately altered in cytokine-activated endothelial cells, despite a >10-fold increase in activity of the sialyltransferase ST6Gal-I that attaches 2–6 linked sialic acids (Hanasaki et al., 1994). Thus, to accurately reveal the glycosylation potential of a given tissue or cell type, there appears no viable alternative to the extraction and direct glycoprofiling of the expressed glycans.
Studies of nucleic acids and proteins can now be done at a whole cell or whole organ level, with gene-chip arrays and 2-D gel based proteomics (Page et al., 1999) providing a global overview of the expression of the genome and the status of the "proteome." Here we have made some early steps towards what might be eventually called "glycomics," i.e., elucidating the complete repertoire of glycan structures in a given cell-type or tissue. Obviously, much work is first needed to resolve the practical limitations in the protocol described here and in effectively profiling the resulting glycan fractions. It is possible that at least some of the steps could be automated. The approach might also be appropriately adapted or modified for other types of biological samples, e.g., insects, plants, and microbes. The potential applications of this approach are many, including the comparison of tissues between populations of a single species, between closely related species, between normal and disease states (e.g., cancer, inflammation), between cell lines with changes in glycosylation, in exploring genetic diseases of humans involving glycosylation and following genetic manipulation of glycosylation in intact cells or animals. We have demonstrated here two examples of the last situation. In each case, we were able to obtain the sought-after information simply by HPAEC-PAD profiling and/or nano-1H-NMR spectroscopy of whole mixtures of N- and O-glycans. In the first case (ST6Gal-I deficiency) we show an essentially complete absence of the enzyme product in the N- and O-glycopeptides from several tissues. In the other case (polypeptide GalNAcT-8 deficiency), we saw little loss of expression of mucin-type O-glycans, corroborating that there are several other enzymes capable of synthesizing the GalNAc-O-Ser/Thr linkage. The clear biological phenotype of the ST6Gal-I null mouse (Hennet et al., 1998), and the lack of an obvious phenotype in the other case (Hennet et al., 1995) now make more sense.
|
Materials and methods |
|---|
|
TopAbstractIntroductionResults and discussionMaterials and methodsAcknowledgmentsAbbreviationsReferences |
|---|
Materials
Most materials were obtained from Sigma Chemical Company. The following materials were obtained from the sources indicated: Sep-Pak C18 cartridges (#54955), Waters–Division of Millipore; SpectraPor membrane dialysis tubing, Spectrum; DEAE-Sephacel and Sephacryl S-200 High Resolution, Pharmacia; Bio-Gel P-2 and sodium dodecyl sulfate (SDS), Bio-Rad; Centricon and Centriprep concentrators, Amicon; ELISA plates (#25801; 96-well flat bottom), Corning; Silica-Gel 60 glass-backed HPTLC plates, Merck; agarose bound Jacalin and endoglycoceramidase, Vector Laboratories Inc.; [6-3H]glucosamine hydrochloride (60 Ci/mmol), American Radiolabeled Chemicals; NaB[3H]4, New England Nuclear (NEN); carrier-free [35S]sodium sulfate, ICN; orcinol, Eastman; ammonium acetate, Fisher; trifluoroacetic acid (TFA), Pierce; chondroitinase ABC, keratanase (Pseudomonas sp. Lyo), and keratanase II (Bacillus sp.), Seikagaku Corp.; and fungal Proteinase K, Gibco BRL. All other chemicals were of reagent grade or better, and were obtained from commercial sources.
Radiolabeled standards
Radiolabeled variant surface glycoprotein labeled in the GPI-anchors by [3H]ethanolamine was a kind gift from Kojo Mensa-Wilmot (University of Georgia) (Mensa-Wilmot et al., 1994). Synthetic (O-GlcNAc-)YSDSPSTST (Reason et al., 1991) with a ß-linked [6-3H]Gal residue attached to GlcNAc was kindly provided by Bradley Hayes and Gerald Hart (Johns Hopkins University). CMP-[9-3H]sialic acid and lactose were incubated with recombinant soluble 2–3 sialyltransferase (ST3Gal-III, kindly provided by Eric Sjoberg, Cytel Corp., San Diego) to enzymatically generate 3'[3H]sialyllactose, and the product was isolated by gel filtration essentially as described previously (Beyer et al., 1981).
Fetuin was radiolabeled in its Sia moieties by mild periodate oxidation followed by reduction with NaB[3H]4 (Van Lenten and Ashwell, 1971). Briefly, bovine fetuin (100 µg) was dissolved in a minimal volume of freshly made 2 mM sodium metaperiodate in phosphate-buffered saline (PBS) and allowed to react on ice, in the dark, for 15 min. A 100-fold molar excess of glycerol was added, and incubation continued for 2 h. The oxidized fetuin was dialyzed (MWCO 1000) against water, recovered by lyophilization, dissolved in 10–20 l 10 mM NaOH and a 4-fold molar excess of NaB[3H]4 in dimethyl formamide (10 µl) was added. After 1–3 h at RT, reduction was completed with a 20-fold molar excess of NaBH4. Excess reducing agent was quenched with acetone (10-fold molar excess relative to NaBH4) for 2 h at RT. The labeled fetuin was reisolated from the void volume of a Bio-Gel P-2 column eluted with 50 mM sodium acetate (pH 5.5), dialyzed (MWCO 1000) against 50 mM sodium acetate (pH 5.5) and then against distilled water. The sample was lyophilized and kept at –20°C until use. The uniform radiolabeling of all N- and O- sialyloligosaccharides was confirmed by releasing them via automated hydrazinolysis (see below), followed by HPAEC monitored by scintillation counting. The profile obtained was similar to that of larger amounts of unlabeled fetuin subjected to the same process and detected by PAD.
Disialoganglioside GD1a (Matreya) was radiolabeled in its Sia moieties as described above for fetuin, except that purification from by-products was done by repeated evaporation with acidified methanol, followed by reverse-phase chromatography on a Sep-Pak C18 cartridge. The sample was loaded in water, the cartridge was washed with water and methanol, and [3H]GD1a eluted with chloroform. Further purification was achieved by HPLC on an Iatrobead 6RS 8010 (10 µm) column (4.6 x 500 mm), eluted with a linear gradient consisting of chloroform/methanol/water from 53:42:5 to 48:42:10 (v/v) over 60 min. The double peak (corresponding to two different acyl chain lengths of the sphingosine moiety) was pooled, dried, dissolved in water, dialyzed against water (MWCO 3500) at 4°C overnight, lyophilized, and kept at –20°C until use.
Tritium and 35S-labeled HS- and CS-GAG glycopeptides were prepared by a modification of a previously described method (Bame and Esko, 1989). Briefly, CHO-K1 cell monolayers (4 P-100 plates) were metabolically radiolabeled with [6-3H]GlcN (100 µCi/plate) in glucose-depleted media (glucose at 0.1 g/l final—10% of normal concentration) for 4 days, or with [35S]sulfate (100 µCi/plate) in sulfate-depleted media with normal glucose concentration (1 g/l), 10 mg/l of cysteine (1/10 of normal concentration), and 3 mg/l methionine (1/5 of normal concentration) (Norgard-Sumnicht and Varki, 1995). Serum (10%) was present during the labeling (dialyzed in the case of the sulfate-depleted media). Labeled cells were washed three times with cold PBS, solubilized with 1 ml per plate of 100 mM NaOH at RT for 20 min, the extract was combined with the original labeling media, and the pH adjusted to 7.0. Proteinase K was added at a final concentration of 2 mg/ml and the mixture incubated at 50°C overnight. The samples were then diluted with water to a salt concentration of 100 mM or lower (checked with a Vapor Pressure Osmometer, 5500XR, Wescor, calibrated against standard solutions). The sample was applied to a 2 ml DEAE-Sephacel column and washed with 30 column volumes of 100 mM NaCl, 20 mM Tris–HCl, pH 7.4. The HS- and CS-GAGs were eluted with 4 column volumes of 1.0 M NaCl, 20 mM Tris–HCl, pH 7.4, and dialyzed against water in MWCO 3500 dialysis tubing.
Tritiated mucin was prepared from one P-100 plate of LS180 human colon carcinoma cells metabolically radiolabeled in Dulbecco’s Modified Eagle’s Medium with 20% fetal calf serum for 3 days with 1 mCi of [6-3H]GlcN. The media containing shed mucins and GAGs was treated with a mixture of chondroitinase ABC (200 mU) and heparinase II (1.25 U) in 50 mM Tris acetate, 2.5 mM CaCl2, pH 6.9 overnight at 37°C. The sample was adjusted to 100 mM NaCl and then loaded onto a 2 ml Jacalin agarose column at 4°C, and the flow was stopped overnight. The column was then washed with 20 ml of 10 mM Tris–HCl, 100 mM NaCl, 1 mM CaCl2, 1 mM MgCl2 (pH 7.5), and bound O-glycosylproteins eluted with 4 column volumes of 100 mM melibiose in the same buffer. The eluate was repeatedly dialyzed against water (MWCO 3500) at 4°C, lyophilized, dissolved in 100 µl of 100 mM Tris–HCl buffer (pH 7.4) and treated with Proteinase K (final concentration 2 mg/ml) overnight at 50°C. After boiling for 10 min, mucin-type glycopeptides were fractionated on a Sephacryl S-200 column (10 ml) in 10 mM ammonium acetate. The largest mucin glycopeptides running in the void volume were pooled and used as a mucin standard.
Non-protein-bound GPI (GIPL) from Trypanosoma cruzi (Y-strain) (Carreira et al., 1996) was a kind gift from Lucia Mendonca-Previato (Universidade Federal do Rio de Janeiro, Brazil). The GIPL (200 µg) was treated with 20 mM metaperiodate in 100 mM sodium acetate buffer (pH 5.5) for 10 min on ice to oxidize between C-5 and C-6 of the galactofuranosyl moiety. The excess reagent was consumed by adding ethylene glycol. After 30 min, the reaction mixture was desalted by microdialysis (