Targeted glycoproteomics: Serial lectin affinity chromatography in the selection of O-glycosylation sites on proteins from the human blood proteome

Journal of Chromatography A, 1132 (2006) 165–173

Targeted glycoproteomics: Serial lectin affinity chromatography in the selection of O-glycosylation sites on proteins from the human blood proteome夽 Malaika Durham ∗ , Fred E. Regnier Department of Chemistry and Cancer Center, Purdue University, West Lafayette, IN 47907, United States Received 14 July 2005; received in revised form 17 April 2006; accepted 28 July 2006 Available online 21 August 2006

Abstract Although lectin selection is gaining increasing acceptance as a tool for targeting glycosylation in glycoproteomics, most of the work has been directed at N-glycosylation. The work reported here focuses on the use of lectins in the study of O-glycosylation. The problem with using lectins for studying O-glycosylation is that they are not sufficiently specific. This paper reports that through the use of serial lectin affinity chromatography (SLAC) it is possible to select predominantly O-glycosylated peptides from tryptic digests of human serum. Jacalin is relatively specific for Oglycosylation but has the problem that it also selects high mannose N-type glycans. This problem was addressed by using a concanavalin A affinity column to first remove high mannose, hybrid-type and biantennary complex-type N-type glycans before application of the Jacalin columns. When used in a serial format, concanavalin A and Jacalin together provide essentially O-glycosylated peptides. The glycoprotein parents of glycopeptides were identified by deglycosylating the selected O-glycopeptides by oxidative elimination. These peptides were then separated by RPC and further analyzed using ESI-MS/MS and MALDI-MS/MS. Using this approach all the O-glycosylated sites in a model protein (fetuin) and over thirty glycoprotein parents from human serum were identified. It is concluded that a serial combination of Con A and Jacalin can be of utility in the study of O-glycosylation in glycoproteomics. © 2006 Published by Elsevier B.V. Keywords: Proteomics; Serial lectin affinity chromatography; O-Glycosylation; Mass spectrometry

1. Introduction Among more than 100 types of post-translational modifications, glycosylation is the most common. It is for this reason that the development of proteomic methods for the study of post-translational modifications by glycosylation is becoming increasingly important [1]. Glycosylation of the N-type on asparagines and O-type on serine or threonine is thought to occur in over 50% of all proteins [2]. The fact that multiple glycan structures and even aberrations in glycosylation can be found at any particular site in a protein complicates glycoproteomics [3]. Alzheimer’s disease [4], certain types of heart disease [5], respiratory illnesses [6], diabetes [7], stress [8], some autoimmune 夽

Presented at the 29th International Symposium on High Performance Liquid Phase Separations and Related Techniques, Stockholm, Sweden, 26–30 June 2005. ∗ Corresponding author. Tel.: +1 314 289 8496x4361. E-mail address: [email protected] (M. Durham). 0021-9673/$ – see front matter © 2006 Published by Elsevier B.V. doi:10.1016/j.chroma.2006.07.070

diseases [9], cystic fibrosis [10], some renal function diseases [11], arthritis [12], cancer [6] and cellular adhesion related diseases have all been associated with aberrations in glycosylation [13]. One of the problems in studying these diseases is in recognizing which glycoform on a specific protein is associated with the disease. Although O-glycosylation plays a major role in regulatory biology, the development of glycoproteomics methods for the study of O-glycosylation is under represented compared to Nglycosylation. The mapping of O-glycosylated proteins and peptides has been achieved in the past by enzymatically tagging proteins and peptides with radio labeled galactose [14]. In addition to the health issues, related to the handling of radio labeled substances, this process can be very lengthy and tedious. Mass spectrometry studies of ␤-eliminated O-linked glycopeptides are a promising new approach, but only model peptides have been examined [15,16]. There have been several O-glycomics studies pioneered by Bertozzi and co-workers, but this requires the introduction of an azide modified monosaccharide to the cell [17–19].

166

M. Durham, F.E. Regnier / J. Chromatogr. A 1132 (2006) 165–173

To date no one has successfully mapped O-glycosylation of a complex protein mixture to the level of specific sites in proteins. Affinity selection followed by deglycosylation of glycoproteins and glycopeptides is being used increasingly in glycoproteomics, particularly in the case of N-glycoproteins. The ready availability of lectins that select specific classes of N-glycans has been widely exploited in the characterization of N-glycosylation In contrast; no single lectin exclusively targets O-glycosylated peptides and proteins. Jacalin is attractive as a selector of O-glycosylated proteins and peptides because it is specific for the GalNAc core found in O-glycosylation. Moreover, binding is not inhibited by the presence of sialic acid in glycans as seen in studies of human cytomegalovirus (CMV) glycoproteins [20–22]. Jacalin is limited by the fact that it also binds high mannose N-linked glycopeptides and the GalNAc core is not found in all O-linked core structures. It will be shown below that the cross-reaction of Jacalin with some N-type glycan structures can be overcome in the selection of O-glycans by first removing these glycans with concanavalin A (Con A). Con A is a broad selectivity lectin that targets the mannose core of N-glycans. It also binds hybrid-type and biantennary complex-type N-glycans to a lesser extent but does not bind more highly branched complex-type N-glycans and O-glycans. Wheat germ agglutinin (WGA) and peanut agglutinin (PNA) are also commonly used to select O-glycans but are not specific for O-glycosylation alone. WGA is specific for the terminal ␣N-acetylglucosamine found in many types of O-glycans [23], but O-GlcNAc is not found in the majority of O-linked core structures. Another limitation is that GlcNAc is also found in the core of N-glycosylated polypeptides. Still another problem with WGA is that the presence of sialic acid can inhibit its binding. PNA is specific for Gal-(␤,1-3)-GalNAc, the common O-glycan core structure, but the presence of sialic acid can prevent binding [24]. Glycopeptides typically have a large number of glycoforms at a single glycosylation site. For both N- and O-glycopeptides, these glycoforms then to coelute from the reverse phase chromatography system together making identification using LC–MS nearly impossible and hinders ones ability to simultaneous sequencing of both the peptide and glycan portions. This means that glycopeptides must be deglycosylated before sequencing either the peptide or glycan moieties. Another problem in the study of O-glycosylation is that no single enzyme is capable of releasing all O-linked glycans from polypeptides. This problem is amplified by the fact that the specificity of the available enzymes is too narrow [25,26] as seen with ␤-galactosidase. Chemical deglycosylation methods are more commonly employed when removing the carbohydrate structure from O-linked glycans. Trifluoromethansulfonic acid is a common chemical method for releasing both N-and O-linked carbohydrates, but lacks the requisite selectivity to remove only O-linked glycans. Deglycosylation with hydrazinolysis is also useful but is associated with degradation of the polypeptide backbone [27–30]. ␤-elimination is a common deglycosylation procedure for O-linked glycans, but amino acid residues are

modified in the process [31]. Glycan elimination with periodate oxidation is a better procedure when there is no need to recover the glycan. The intent of this work was to develop and evaluate a separation method that when coupled with current and future advances in mass spectrometry based sequencing could be used for examining O-glycosylation in glycoproteins. The focus of the work described here was on the development of a broad affinity based method for the selection and identification of peptides with Oglycosylation sites in a proteome level study of human serum samples. 2. Materials Monobasic potassium phosphate, sodium borohydride (96%), sodium chloride, sodium hydroxide, sodium phosphate, sodium acetate and HPLC-grade acetonitrile were all purchased from Mallinckrodt Baker (Phillipsburg, NJ, USA). Sodium m-periodate, sodium azide, ␣-cyano-4-hydroxycinnamic acid, 4-(2-hydroxyethyl)piperazine-1-ethanesulfonic acid (HEPES), neuraminidase, lectin from Canavalia ensiformis (Concanavalin A), bovine fetuin, 2-(dimethylamino) ethanethiol hydrochloride 95%, human sera, glycine, glucose, sodium cyano borohydride, iodoacetic acid, urea, l-cysteine, N-hydroxysuccinamide, acetic acid N-hyrdroxysuccinamide ester (acetic anhydride)-d6 , trypsin and tosyl lysine chloroketone (TLCK) trypsin inhibitor were all purchased from Sigma–Aldrich (St. Louis, MO, USA). Lectin from Artocarpus integrifolia (Jacalin) was purchased from Vector Laboratories (Burlingame, CA, USA). Tris buffered saline (TBS) and methyl alpha-d-mannopyranoside in TBS were purchased from EY Laboratories Inc. (San Mateo, CA, USA). Calcium chloride and sodium acetate were both purchased from Fisher (Fairlawn, NJ, USA). Dithiothreitol (DTT), Cleland’s reagent, and HPLC grade trifluoroacetic acid (TFA) were purchased from Pierce (Rockford, IL, USA). LiChrospher 1000 diol silica was purchased from EM Science (Darmstadt, Germany). Double-deionized water (ddI H2 O) was produced by a Milli-Q gradient A10 system from Millipore (Bedford, MA, USA). All commercially available reagents were used without purification. 3. Methods 3.1. Lectin column synthesis A. integrifolia lectin (Jacalin) was immobilized on silica using a modified procedure from Borreback et al. [32]. Two and one half grams of LiChrospher 1000 Diol silica was oxidized for 2 h with agitation at room temperature in 50 mL of 9:1 acetic acid: H2 O (v/v) containing 2.5 g of sodium m-periodate. The solution was centrifuged and the collected particles washed with 50 mL H2 O. The silica particles were then suspended in 5 mL of 0.1 M Na2 PO4 (pH 7) and sonicated for 5 min. Twenty five milligrams of Jacalin was dissolved in 5 mL of 0.1 M Na2 PO4 (pH 7) and added to the silica particle suspension. NaCNBH3 (72.5 mg) in 1.25 mL of the same phosphate buffer was added to the mixture and incubated at 4 ◦ C overnight. One hundred milligrams of solid NaBH4 were added to the reaction vessel and

M. Durham, F.E. Regnier / J. Chromatogr. A 1132 (2006) 165–173

the mixture incubated another 30 min at room temperature. The Jacalin affinity sorbent was isolated from the reaction mixture by filtration and washed extensively with 0.15 NaCl in 20 mM Na2 PO4 at pH 7. The affinity sorbent was then packed into a 4.6 mm × 100 mm PEEK column. After the Jacalin affinity sorbent was packed into the PEEK column, the binding capacity of the silica was determined by using a Varian Cary 300 UV–vis spectrophotometer (Palo Alto, CA, USA). The remaining solution was run on the UV–vis spectrophotometer and the wavelength was monitored at 277 nm. Based on a calibration curve of several standards, it was determined that 7.62 mg of Jacalin was bound to a gram of silica. 3.2. Proteolysis Trypsin was used in the proteolysis of bovine fetuin and human sera. Five hundred milligrams of bovine fetuin were added to 6 M urea and 10 mM DTT in 4 mL of 50 mM HEPES and 10 mM CaCl2 . The mixture was incubated at room temperature for 2 h. Forty millimolars iodoacetic acid was added and the mixture incubated in darkness on ice for 2 h. Twenty millimolar l-cysteine was added to quench the alkylation reaction. Using a solution of 10 mM CaCl2 in 50 mM HEPES buffer the mixture was diluted to a final urea concentration of 1.15 M. Trypsin was added at a 1:50 weight ratio of enzyme to total protein. The solution was then incubated in a water bath at 37 ◦ C overnight. After proteolysis, TLCK trypsin inhibitor was added at a molar concentration that exceeded trypsin by two-fold. Neuraminidase (1 unit) was added to the digested proteins and allowed to incubate for 30 min in a water bath at 37 ◦ C to remove sialic acid residues on peptides. The removal of the sialic acids was carried

167

out to ensure that there was no interaction between these groups and any of the lectins used. The digest was then stored in 50 mL polypropylene tubes at 0 ◦ C until needed for chromatography. For the proteolysis of human sera, 4 mL of the sera was added to 6 M urea and 10 mM DTT in 4 mL of 50 mM HEPES and 10 mM CaCl2 . All other steps were the same. 3.3. Selection of O-linked glycopeptides In addition to being able to select O-linked glycopeptides, Jacalin can also bind N-linked high-mannose glycopeptides. High-mannose glycopeptides were removed from the tryptic digest with concanavalin A (Con A) agarose. Con A binds to high mannose N-linked glycopeptides. Using a Biocad Microanalytical Workstation from PE Biosystems (Farmingham, MA, USA), the Jacalin column was equilibrated with 0.01 M phosphate buffered saline (PBS) (pH 7.45) using a flow rate of 0.33 mL/min. Two milliliter of Con A “cleaned” sample (fetuin or human sera) was applied to the Jacalin column. After eluting unbound peptides, the O-linked glycopeptides were eluted with 0.8 M galactose. The eluent was monitored at 277 nm. 3.4. Synthesis of d3 -N-acetoxysuccinamide The eluted O-linked glycopeptides from fetuin and human sera were labeled with either N-acetoxysuccinamide or d3 labeled N-acetoxysuccinamide with the deuterium on the methyl group of the acetate moiety. Samples were labeled in order to aid in the identification of O-linked glycopeptides and to evaluate the method efficiency in a modification of published methods [33].

Fig. 1. (a) Oxidative elimination, (b) Michael addition to glycan removed from O-glycosylated peptide (NeuAca2-6GalNAc-O).

168

M. Durham, F.E. Regnier / J. Chromatogr. A 1132 (2006) 165–173

A 100 mL round bottom flask and stir bar were dried under nitrogen. N-hydroxysuccinamide (1.77 g) was added to the flask under nitrogen. Five grams of (acetic anhydride)-d6 , in its liquid form, were then added to the flask. The solution was stirred overnight and then filtered and washed with hexane. The crystals were dried and kept at 0 ◦ C until needed. 3.5. Deglycosylation The eluted O-linked glycopeptides were deglycosylated before further fractionation (Fig. 1). A modification of the oxidative elimination process described by Hong and Kim [34].was used to remove the O-linked glycopeptides. Three milliliters of 100 mM NaOH was added to 6 mL of eluent and allowed to incubate for 30 min at room temperature. After incubation, 100 mmole of NaIO4 in 3 mL 100 mM sodium acetate buffer (pH 4.5) was added and the mixture allowed to incubate overnight at 4 ◦ C. One milliliter of 2% glycine was added to the eluent mixture and allowed to incubate at room temperature for 30 min. Fifty milligrams of 2-(dimethylamino) ethanethiol hydrochloride was then added. Oxidative elimination of glycans was achieved by adding 100 mmole of NaOH. When a reduction occurs on the carbohydrate moiety 2-(dimethylamino) ethanethiol reacts with the newly formed double bond in the cleaved sugar [35]. This Michael addition prevents the side reactions with free amines. The deglycosylated peptides were stored at 0 ◦ C. 3.6. Chromatography All chromatographic separations were performed on a Biocad Microanalytical Workstation from PE Biosystems (Farmingham, MA, USA). The deglycosylated peptide mixture was applied to a 4.6 mm × 250 mm C18 column. The deglycosylated peptide fraction was eluted from the reversed phase column at a flow rate of 1.0 mL/min in a 60 min linear gradient using 5% acetonitrile in 0.1% aqueous TFA (Buffer A) to 60% acetonitrile in 0.1% aqueous TFA (Buffer B). Eluted peptides were monitored at 215 nm and the fractions were analyzed by off-line ESI-MS, ESI-MS/MS and MALDI-MS/MS mass spectrometry. 3.7. ESI mass spectrometry Electrospray ionization mass spectrometry was performed on a PE SCIEX QSTAR hybrid Q-TOF MS (Boston, MA, USA). MS–MS sequencing was achieved by selecting the parent ion in the quadrupole using the low-resolution mode. This method of selection ensures that the entire isotope cluster is introduced into the collision cell. 3.8. MALDI mass spectrometry Matrix-assisted laser desorption/ionization mass spectroscopy was performed on an Applied Biosystems VoyagerDETM 4700 MALDI-TOF-TOF Workstation.

3.9. Data analysis Spectral peaks produced in the ESI positive ion mode arise from multiply charged molecular ions. Peak products in MALDI-MS/MS correspond to singly charged molecular ions. The ESI-MS spectra from bovine fetuin were analyzed manually using the SWISS-PROT database while computer-assisted searching with MASCOT was used to interpret the ESI-MS/MS and MALDI-MS/MS spectra of the deglycosylated peptides from human serum. The following search parameters were used with the MASCOT program; Database: NCBInr, Enzyme: TrypChymo, Missed Cleavages: 4, Variable Modifications: Carboxymethyl (C), Sodiated (C-term), Sodiated (DE), Acetyl heavy (K), or Acetyl light (K), Fixed Modification: Acetyl heavy (N-term) or Acetyl light (N-term). The enzyme selected was TrypChymo because trypsin sometimes cleaves in a manner similar to chymotrypsin. Since the MS is performed off-line and the fractions collected from the reverse phase system have to be reconstituted, there may be a presence of sodium in the samples. 4. Results and discussion 4.1. Analytical strategy The focus of this study was to select and identify O-linked glycopeptides from a model protein and human serum. The analytical strategy focused on the use of lectins in series. Mannose rich glycopeptides were pre-selected with Con A before the sample came in contact with the Jacalin column. Con A is widely used to bind high mannose N-glycosylated peptides [36,37]. The selected O-linked glycopeptides were chemically deglycosylated and further fractionated using reversed phase chromatography. Identification of the deglycosylated peptides was achieved by, ESI-MS, ESI-MS/MS and MALDI-MS/MS mass spectrometry. A modified periodate oxidative method was used in this study to cleave the bond between the C3 and C4 carbon atoms in GalNAc and render the resulting oxidized glycan susceptible to elimination [39]. The net result is that all oxidized O-linked carbohydrate residues are removed from the serine or threonine residues in peptides to which they were coupled. This method is distinctly different than the common ␤-elimination of O-linked glycans from polypeptides in that the double bond is formed on the glycan, not the peptide. The solution with the double bond containing glycans is treated with 2-(dimethylamino) ethanethiol, causing a Michael addition to the double bond. This prevents any side reactions with primary amines of peptides in the solution. Deglycosylation of complex mixtures of glycopeptides generates two sets of products, those arising from the peptide portion of the conjugate and the other from the glycan portion. To assist in the differentiation between these two moieties, primary amine groups on the peptide were stable isotope coded. Dividing the samples into two fractions and differentially labeling the amine groups in the two halves with acetate and trideuteroacetate,

M. Durham, F.E. Regnier / J. Chromatogr. A 1132 (2006) 165–173

respectively, provided the coding. Peptides thus, coded appeared as doublets in mass spectra separated by 3–6 amu, depending on the number of amines in the peptide whereas products from glycan cleavage appeared as singlets. This labeling strategy also allowed examination for the efficiency of the method. Since the SLAC selected peptides were divided into two fractions, labeled and mixed back together in a 1:1 ratio all the spectra peaks arising from peptides should appear as the same height. By analyzing several peptides from several different runs, it was found that on average the peptides were appearing in a 1:1.150 ratio with a standard deviation of 0.114. The analytical scheme can be see in Fig. 2. It can be argued that because only the O-linked glycans were cleaved from the peptides, there would be no need for Con A selection to remove N-linked glycan containing peptides. That is true in the studies to be described below. In future studies, it is the intent to carry out CID and ETD sequencing studies of glycopeptides. It would be important in such studies that no N-linked glycopeptides were present in samples. 4.2. Fetuin analysis The method described above was validated using bovine fetuin. Fetuin is a good model protein because it is only 38 kDa, is readily available from fetal calf serum, and contains both N-linked and O-linked glycosylation sites. The O-linked glycosylation sites are at serine residues 271, 280, 341 and on threonine 282. Five hundred milligrams of fetuin was tryptic digested and desialylated. Two milliliter of the fetuin digest were then run on the C18 reverse phase column (Fig. 3a). The resulting chro-

Fig. 2. Analytical scheme.

169

matogram is complex, characteristic of a tryptic digested protein. A 2 mL aliquot of the fetuin digest containing 80 ␮g of protein was applied to the Con A agarose column in order to remove N-linked glycosylation, especially the high mannose type, which may bind to Jacalin. Glycopeptides not retained by the Con A column were then applied to the Jacalin column to select Gal-␣(1,3)-GalNAc containing O-linked glycopeptides. The selected glycopeptides were acetate labeled and then fractionated using reverse phase chromatography. Because of the small number of O-linked glycosylation sites found in fetuin, the complexity of the chromatogram was greatly reduced after Jacalin selection (Fig. 3b). But the chromatogram is still more complex than would be expected for a protein with only four glycosylation sites. Another feature of this chromatogram and that in Fig. 3c is the apparent low-resolution of the separation. The broad peaks seen in the chromatogram would suggest poor resolution or the need for optimization. That is the nature of chromatograms with hundreds of components. Actually, the RPC column was new and gave a very good separation of a bovine serum albumin tryptic digest (data not shown). The glycan heterogeneity problem noted above was address by oxidative elimination of glycans from peptides before MS identification of peptide sequences. After deglycosylation the peptides were further fractionated using reverse phase chromatography (Fig. 3c). As judged from the chromatograms in Fig. 3b and c, sample complexity was reduced slightly by deglycosylation but the chromatogram was still more complex than expected. All the known O-glycosylation sites in fetuin were identified by ESI/MS and use of the SWISS-PROT database (Table 1). The identification was achieved manually, looking for masses that corresponded to peptides with glycosylation sites. Serial lectin affinity chromatography (SLAC) of glycopeptides from fetuin with tandem Con A and Jacalin columns appears to select O-glycosylated peptides, as expected, and is a simple way to target O-glycosylation. There are, however, some features of the method that merit discussion. One is the amount of sequence redundancy among the selected peptides. The glycosylation site at amino acid 341 was found in 12 peptides while the glycosylation sites at amino acids 271, 280 and 282 were found in 3 peptides. Clearly the presence of O-glycans on residue 341 of fetuin affected the rate of trypsin digestion and resulted in partial proteolysis. This accounts for larger than expected number of peaks in the reversed phase chromatograms in Fig. 3b and c. Increasing the digestion time or using an immobilized trypsin column with a much higher enzyme to protein ratio would diminish this phenomenon. There is also a large number of miscleaveges in many of the peptide chains. This is most likely due to the fact that the large carbohydrate groups present in fetuin keep trypsin from cleaving the protein efficiently. The second feature of the SLAC process worth noting is that some peptides were captured non-specifically. The data in Table 1 shows that 5 peptides from fetuin without glycosylation sites were captured by the SLAC method. This could be due to the immobilization of Jacalin on a silica-based support. It has been shown that non-specific binding of peptides on a Con

170

M. Durham, F.E. Regnier / J. Chromatogr. A 1132 (2006) 165–173

Fig. 3. (a) Reverse-phase chromatogram of acetate labeled fetuin peptides after tryptic digestion and desialydation. (b) Reverse-phase chromatogram of acetate labeled fetuin peptides after selection by SLAC. (c) Reverse-phase chromatogram of deglycosylated acetate labeled fetuin peptides after selection by SLAC. Peptides were eluted from a 250 mm × 4.6 mm I.D. C18 column with a 60 min linear gradient ranging from 5% ACN in a 0.1% aqueous TFA to 60% ACN in 0.1% aqueous TFA at a flow rate of 1 mL/min.

A/silica column could be eliminated by immobilizing Con A on sepharose instead of silica [38]. Perhaps the same would be true here as well. The silica matrix was chosen for immobilizing Jacalin over purchasing agarose bound Jacalin for several

reasons. One is that the soft agarose matrix is difficult to use in HPLC systems. The second is that the concentration of Jacalin immobilized on silica is much higher than the commercially available agarose bound Jacalin.

Table 1 Peptides from fetuin Fraction (#)

Sequence

Carbohydrate

1 3 4 6 7 8 9 10 11 12 13 14 15 16 18 19 20 21

QDGQF SGVASVESSSGEAFHVGKTPIVGQPSIPGGPVRLCPGRIRY TPIVGQPSIPGGPVRLCPGRIRY QQTQHAVEGDCDIHVLK HVGKTPIVGQPSIPGGPVRLCPGRIRYF TPIVGQPSIPGGPVRLCPGRIRYF HTFSGVASVESSSGEAFHVGKTPIVGQPSIPGGPVRLCPGR HVGKTPIVGQPSIPGGPVRLCPGRIR SGVASVESSSGEAFHVGKTPIVGQPSIPGGPVR QTQPVIPQPQPDGAEAEAPSAVPDAAGPTPSAAGPPVASVVVGPSVVAVPLPLHR DLRHTFSGVASVESSSGEAFHVGKTPIVGQPSIPGGPVR VPLPVSVSVEFAVAATDCIAKEVVDPTK GSVIQKALGGEDVRVTCTLFQTQPVIPQPQPDGAEAEAPSAVPDAAGPTPSAAGPPVASVVVGPSVVAVPLPLHRAHY HVGKTPIVGQPSIPGGPVRLCPGR ALGGEDVRVTCTLFQTQPVIPQPQPDGAEAEAPSAVPDAAGPTPSAAGPPVASVVVGPSVVAVPLPLHRAHYDLR HVGKTPIVGQPSIPGGPVRLCPGRIR TPIVGQPSIPGGPVRLCPGRIRY LQLVEISRAQF

None 341 341 None 341 341 341 341 341 271, 280, 282 341 None 271, 280, 282 341 271, 280, 282 341 341 None

Bolded and underscored amino acids correspond to O-glycosylation sites.

M. Durham, F.E. Regnier / J. Chromatogr. A 1132 (2006) 165–173

4.3. Analysis of human serum Human serum is a complex mixture containing thousands of proteins, many with either N- or O-linked glycosylation, or both. Following serum proteolysis sample complexity increases at least 50-fold [39]. The great advantage of lectin affinity chromatography is that it provides an efficient method for reducing this enormous sample complexity [5]. Human serum was tryptic digested, desialyated, lectin affinity selected, and acetate labeled in the same manner as the fetuin sample. The chromatogram (Fig. 4a) is complex, probably due to a combination of the presence of many O-linked glycoproteins in serum, glycan heterogeneity on peptides, and incomplete

171

proteolysis. The heterogeneity problem was again addressed by deglycosylating the O-linked glycopeptides after lectin selection. The selected peptides were then further fractionated with a C18 reverse phase column. As in the case of fetuin, when the SLAC selected, deglycosylated sample was resolved using the reverse phase system there was a reduction in the number of peaks versus the glycosylated sample. This suggests that reversed phase column resolved some of the glycoforms of peptides (Fig. 4b). The deglycosylated peptides were characterized and identified by using ESI-MS/MS and MALDI-MS/MS technology. The limit of detection with this method is in the fmolar range, as expected. The recovery of peptides from lectin columns is high and peptides are being iden-

Table 2 Parent proteins from human serum Protein

Sequence

C3HC4-type zinc finger protein Coagulation factor XIII A chain precursor Copine VIII Erythrocyte membrane glycoprotein Rh50 Glucocorticoid-induced leucine zipper protein Ig heavy chain V-III region (ART) Ig heavy chain variable region, VH3 family Immunoglobulin Immunoglobulin domain cell adhesion molecule (cam) subfamily Immunoglobulin domain variable region (v) subfamily Immunoglobulin E variable region

LEQLCEEFSEEERVR AEVNSDLIYITAKKDGTHVVENVDATHIGK DSRYNSTAGIGDLNQLSAAIPATR RIHDTCGVHNLHGLPGVVGGLAGIVAVAMGASNT TLASPEQLEK EPQVYTLPPSR QTPDRGL HDYAPSVK ASSRLRPEAEPELGVKTPEEGCL VVSVLTVLHQDWLDGK GLEWVSGIGDSGGKTY

Immunoglobulin heavy chain variable region

HYEESVR ADGGTIDYGAPLKGR HYEESVR EWVSYISNSDDTRRY EVDVSPK

Immunoglobulin heavy chain VHDJ region Immunoglobulin kappa chain V-J region

ESGPTLVKPTQTVTLTCTFSGFSL DSLPTFGGGTKVDIK

Immunoglobulin kappa light chain variable region

LEINRTVAAPSNF ISGASSLESGVPSRF

Immunoglobulin M chain Immunoglobulin M heavy chain Insulin-Like Growth Factor II Interleukin 2 receptor Interleukin 6 receptor Leucine-rich repeats Lipoprotein Gln I Lutheran blood group glycoprotein precursor Mucin glycoprotein Oxidative 3 alpha hydroxysteroid dehydrogenase Protocadherin beta 6 precursor Rho guanine nucleotide exchange factor 3 Ring finger protein 20

MTPAFVTMTSPCHFTLVTGVHCEVQL DGNYANCHESLPHY YFSRPASR REAGEEVPDAGPREGVSFPWSRPPGQGEFR RTIIGNETAVNVDSSHTEY LVEVERMECATPSDK DEPPQSPWDR SPPYQLDSQGR LSSTGPSPSSNHTPASPTQTPL QVVSHLQDK DLDAGSFGK NAAPSSTK RAAGEPGTSMPPEKK

Semaphorin domain

AIDAVIYRSLGESPTLR HLLACGTGAFQPTCALI TVGHR

Sodium/calcium exchanger protein

ALAIVCDDFFVPSLEK GEYPPDLFSVEER

T cell receptor beta chain Tetratricopeptide repeat protein 7 Thrombospondin N-terminal-like domains Titin Von Willebrand factor (vWF) type C domain Zinc finger, CSL domain containing 2

CASSEAFWTGTEA HASGFLGEHSPGGQR AIWQITDRDYKPQVGVIADSSK ITFVENVATLQFAK HPTIILAQQEAVEGGCSHLGQSY EDLENGEDVATCPSCSLIIKVIYDK

172

M. Durham, F.E. Regnier / J. Chromatogr. A 1132 (2006) 165–173

Fig. 4. (a) Reverse-phase chromatogram of tryptic glycopeptides from human serum after selection by SLAC. (b) Reverse-phase chromatogram of acetate labeled, deglycosylated human serum tryptic peptides after selection by SLAC. Peptides were eluted from a 250 mm × 4.6 mm I.D. C18 column with a 60 min linear gradient ranging from 5% ACN in a 0.1% aqueous TFA to 60% ACN in 0.1% aqueous TFA at a flow rate of 1 mL/min.

tified by RPC/MS methods identical to those being used in other types of proteomics. The spectra from the fragmented samples were fed into the Internet database search program MASCOT MS/MS ions search. This program used the NCBIr database to identify protein parents of peptides. The results are shown in Table 2. All peptides identified had a MASCOT score of 25 or greater. To ensure that no N-glycosylated peptides were identified a combination of methods were used. By using Con A, it was possible to remove any high mannose N-glycosylated peptides that bind to Jacalin. Also the oxidative elimination method that was used only removes glycan attached to a ␤-hydroxyl group. The use of this combination ensures that only O-glycosylated peptides have been selected and identified in Table 2. All of the peptides identified (Table 2) either have a known site of O-glycosylation or contain a domain that is known to be O-glycosylated. Also all the peptides identified are from serum proteins or proteins which are known to be excreted. Examples cited in the literature as being O-glycosylated are the peptide (LSSTGPSPSSNHTPASPTQTPL) from mucin [40,41], (YFSRPASR) from Insulin-like growth factor II (REAGEEVPDAGPREGVSFPWSRPPGQGEFR), from Interleukin 2 receptor and (HPTIILAQQEAVEGGCSHLGQSY) from von Willebrand factor. Abundant proteins were not removed from samples before analysis, so a small number of cases of non-specific binding (NSP) from prominent serum proteins were found but not included in Table 2. It is also interesting that some Oglycosylated proteins were not identified, as in the cases of CD8 and IgM. 5. Conclusions The data presented here lead to the conclusion that in the case of tryptic digests of blood; lectins provide a simple, but rapid method to globally select and enrich glycopeptides from complex samples. Among the most substantial advantages seen were: (1) the selection and enrichment of glycopeptides from complex samples along with; (2) a very sizeable reduction in mixture diversity and; (3) the identification of O-glycosylation sites in proteins. The key new feature of this work was the use of

serial lectin affinity chromatography to increase the selectivity of lectin affinity chromatography columns for glycoproteomics of O-linked oligosaccharides. Although a Jacalin column alone will not exclusively select O-glycosylated polypeptides, removal of N-glycosylated species from a complex mixture with Con A before Jacalin selection greatly improved O-glycopeptide selection specificity from a proteome. This serial selection approach allowed the identification of a broad variety of O-glycosylated proteins. A weakness of the method is that glycans have to be removed from peptides to achieve their identification. This results in the loss of a very substantial amount of structural information that is probably of biological significance. Ideally, it would be possible to sequence both the peptide and glycan portions of a glycopeptide simultaneously. Preliminary studies in which both CID and ECD were applied to glycopeptides indicate it may be possible to sequence both the peptide and glycan portions of glycopeptides in an ion trap mass spectrometer without deglycosylation [42]. O-Glycosylation seems to increase in many kinds of diseases. For this reason it is critical that methods be available for identifying the proteins involved, sites of glycosylation within these proteins, and the actual glycan structures. The significance of the method described here is that it enables the study of Oglycosylation independent of N-glycosylation. Acknowledgement The authors gratefully acknowledge financial support from a National Institute of Health Grant (GM59996) to Purdue University. References [1] L. Wells, K. Vosselle, R. Cole, J. Cronshaw, M. Matunis, G. Hart, Mol. Cell Proteomics 1 (2002) 791. [2] A. Varki, R. Cummings, J. Esko, H. Freeze, G. Hart, J. Marth, Essential of Glycobiology, Cold Spring Harbor Laboratory Press, New York, NY, 1999, p.125. [3] L. Xiong, F. Regnier, J. Chromatogr. B 782 (2002) 405. [4] N. Ahmed, O. Argirov, H. Minhas, C. Cordeiro, P. Thornalley, Biochem. J. 364 (2002) 1.

M. Durham, F.E. Regnier / J. Chromatogr. A 1132 (2006) 165–173 [5] F. Spinale, M. Coker, L. Heung, B. Bond, H. Gunasinghe, T. Etoh, A. Goldberg, J. Zellner, A. Crumbley, Circulation 102 (2000) 1944. [6] E. Puchelle, S. de Bentzmann, J. Zahm Respiration 62 (1995) 2. [7] H. Vlassara, M. Palace, J. Intern. Med. 251 (2002) 87. [8] W. Matis, R. Lavker, G. Murphy, J. Invest. Dermatol. 94 (1990) 492. [9] M. Sato, T. Toyozaki, K. Odaka, T. Uehara, Y. Arano, H. Hasegawa, K. Yoshida, K. Imanaka- Yoshida, T. Yoshida, M. Hiroe, H. Tadokoro, T. Irie, S. Tanada, I. Komuro, Circulation 106 (2002) 1397. [10] S. Kirkham, J. Sheehan, D. Knight, P. Richardson, D. Thornton, Biochem. J. 361 (2002) 537. [11] K. Sebekova, P. Blazicek, D. Syrova, Z. Krivosikova, V. Spustova, A. Heidland, R. Schinzel, Kidney Int. 78 (2001) S58. [12] G. Reddy, S. Dhar, Agents Actions 25 (1988) 63. [13] M. Jaradat, C. Schnizlein-Bick, G. Singh, S. Moe, Kidney Int. 59 (2001) 1951. [14] E. Roquemore, T. Chou, G. Hart, Methods Enzymol. 230 (1994) 443. [15] C. Settineri, K. Medzihradszky, F. Masiarz, A. Burlingame, C. Chu, C. George-Nacimento, Biomed. Environ. Mass Spectrom. 19 (1990) 655. [16] G. Rademaker, J. Haverkamp, J. Thomas-Oates, Org. Mass Spectrom. 28 (1993) 1536. [17] J. Prescher, D. Dube, C. Bertozzi, Nature 430 (2004) 873. [18] D. Vocadlo, H. Hang, K. Eun-Ju, J. Hanover, C. Bertozzi, PNAS 100 (2003) 9116. [19] E. Saxon, C. Bertozzi, Science 287 (2000) 207. [20] R. Theiler, T. Compton, J. Virology 76 (2002) 2890. [21] R. Coppo, L. Peruzzi (Eds.), Moderately Proteinuric IgA Nephropathy in the Young. Biomedical Health Research, ISO Press, Amsterdam, Netherlands, 2000, p. 73. [22] R. Gregory, J. Rundegren, R. Arnold, J. Immunol. Methods 99 (1987) 101.

173

[23] W. Adair, S. Kornfeld, J. Biol. Chem. 249 (1974) 4696. [24] R. Lotan, N. Sharon, Methods Enzymol. 138 (1978) 361. [25] F. Hanisch, G. Uhlenbruck, H.J. Peter-Katalinic, J. Biol. Chem. 264 (1989) 872. [26] P. Rudd, R. Dwek, Curr. Opin. Biotech. 8 (1997) 488. [27] T. Gerken, R. Gupta, N. Jentoft, Biochemisty 31 (1992) 639. [28] A. Zinn, J. Planter, D. Carlson, The Glyconjugates, Academic Press, New York, NY, 1977, p. 82. [29] F. Dwons, W. Pigman, Methods in Carbohydrate Chemistry, Academic Press, New York, NY, 1976. [30] K. Greis, B. Hayes, F. Comer, M. Kirk, S. Barnes, T. Lowary, G. Hart, Anal. Biochem. 234 (1996) 38. [31] S. Brittain, S. Ficarro, A. Brock, E. Peters, E. Nature, Biotechnology 23 (2005) 463. [32] C. Borrebaeck, J. Soares, B. Mattiasson, J. Chromatogr. 284 (1984) 187. [33] K. Jacobson, J. Med. Chem. 28 (1985) 1341. [34] J. Hong, Y. Kim, Glycoconjugate 17 (2000) 691. [35] H. Steen, M. Mann, JASMS 13 (2002) 996. [36] Y. Yasuda, S. Ikeda, S. Sakai, T. Tsukuba, K. Okamoto, K. Nishishita, A. Akamine, Y. Kato, K. Yamamoto, Eur. J. Biochem. 266 (1999) 383. [37] N. Suzuki, Y. Lee, Glycobiology 14 (2004) 275. [38] D. Ren, N. Penner, B. Slentz, H. Mirzaei, F. Regnier, J. Proteome Res. 2 (2003) 321. [39] X. Zhang, Ph.D. Thesis, Purdue University, West Lafayette, IN, 2001. [40] I. Porowska, A. Paskiewicz-Gadek, T. Anchim, S. Woklczynski, A. Gindzienski, Int. J. Mol. Med. 13 (2004) 459. [41] I. Brockhausen, Biochim. Biophys. Acta 1473 (1999) 67. [42] J. Hogan, S. Pitteri, P. Chrisman, S. McLuckey, J. Proteome Res. 4 (2005) 628.