Enhanced N-Glycosylation Site Analysis of Sialoglycopeptides by Strong Cation Exchange Prefractionation Applied to Platelet Plasma Membranes *S

Elucidation of post-translational modifications to proteins, such as glycosylations or phosphorylations, is one of the major issues concerning ongoing proteomics studies. To reduce general sample complexity, a necessary prerequisite is specific enrichment of peptide subsets prior to mass spectrometric sequencing. Regarding analysis of overall N-glycosylation sites in the past, this has been achieved by several approaches proving to be more or less complicated and specific. Here we present a novel strategy to target N-glycosylation sites with application to platelet membrane proteins. Initial aqueous two-phase partitioning for membrane enrichment and single step strong cation exchange-based purification of glycopeptides resulted in identification of 148 glycosylation sites on 79 different protein species. Although 69% of these sites were not annotated in the Swiss-Prot database before, a high number of 75% plasma membrane-localized proteins were analyzed. Furthermore miniaturizations and relative quantification are comprised in the developed method suggesting further use in other proteome projects. Results on platelet glycosylation sites may imply an impact on research of bleeding disorders as well as potential new functions in inflammation and immunoactivity.

Elucidation of post-translational modifications to proteins, such as glycosylations or phosphorylations, is one of the major issues concerning ongoing proteomics studies. To reduce general sample complexity, a necessary prerequisite is specific enrichment of peptide subsets prior to mass spectrometric sequencing. Regarding analysis of overall N-glycosylation sites in the past, this has been achieved by several approaches proving to be more or less complicated and specific. Here we present a novel strategy to target N-glycosylation sites with application to platelet membrane proteins. Initial aqueous two-phase partitioning for membrane enrichment and single step strong cation exchange-based purification of glycopeptides resulted in identification of 148 glycosylation sites on 79 different protein species. Although 69% of these sites were not annotated in the Swiss-Prot database before, a high number of 75% plasma membrane-localized proteins were analyzed. Furthermore miniaturizations and relative quantification are comprised in the developed method suggesting further use in other proteome projects. Results on platelet glycosylation sites may imply an impact on research of bleeding disorders as well as potential new functions in inflammation and immunoactivity. Molecular & Cellular Proteomics 6:1933-1941, 2007.
During recent years, the fields of proteomics and glycomics have experienced rapid advancements both in analytical study design as well as instrumental setup. Besides common proteomics experiments monitoring complete cells or tissues, e.g. by differential two-dimensional PAGE, current focus has turned to subproteomes or specific subsets of proteins. Glycomics has emerged as a field of growing interest due to widespread importance of carbohydrate additions to proteins.
Glycosylations are divided into several classes, such as Nglycosylation, O-glycosylation (1), glycosylphosphatidylinositol anchors, C-mannosylation (2), etc., as well as a high number of structural isoforms. Therefore, many strategies have focused on distinct subsets of glycoproteins or -peptides. To analyze glycan attachment sites, several methods have been reported in the past. Carbohydrate-lectin interactions have been extensively used for isolation of glycopeptides and subsequent site analysis (3). Moreover trapping by hydrazide chemistry (4) as well as enrichment of glycopeptides by normal phase chromatography has been demonstrated (5). As shown with platelets, hydrazide chemistry and lectin affinity approaches are readily applicable to these anucleate cells (6). In this context, former proteome studies on platelets revealed that only a small proportion of membrane-bound proteins are accessible by traditional proteomics techniques such as twodimensional PAGE coupled with mass spectrometry (7)(8)(9). Therefore, subfractionation is a prerequisite to cover the high dynamic range of protein isoforms within platelets. For the identification of N-glycosylation sites especially on platelet membrane proteins we focused on aqueous two-phase partitioning for membrane enrichment and specific purification of a glycopeptide subset by strong cation exchange (SCX) 1 chromatography.
Recently the method of polymer-based two-phase partitioning gained interest in the field of proteomics due to its simplicity and reliability (10 -12). Most commonly polyethylene glycol/dextran mixtures are used to generate a system comprising 1) a polyethylene glycol-rich upper phase and 2) a lower phase enriched with dextran (13). Upon careful choice of parameters such as salt and polymer concentration, plasma membranes tend to partition preferentially to the upper phase due to their physicochemical properties (14). Thereby they are separated from the bulk of intracellular membranes, e.g. endoplasmic reticulum and mitochondria. However, in the case of platelets a strict partitioning is limited due to their unique cellular morphology comprising e.g. a dense tubular membrane system. Nevertheless the method has the advantage of rapid and gentle sample preparation. It can also be easily scaled in respect to available sample amounts. Although membrane preparations are already enriched in glycoproteins, they still contain high numbers of non-glycosylated proteins, e.g. high abundant structural components like actin. A further purification of plasma membrane glycopeptides is necessary to assess N-glycosylation sites in larger scale by avoiding suppression effects during mass spectrometric analysis.
In contrast to already mentioned glycopeptide enrichment methods based on specific trapping of glycopeptide subsets, SCX enrichment increases the relative amount of glycopeptides by trapping and removing all non-glycopeptides in the sample. The principle of this novel approach (see Fig. 1) is based on the assumption that the majority of extracellular N-glycans contains sialic acid residues, which reduce the positive net charge of glycopeptides in comparison with other peptides. At low pH the majority of tryptic peptides are theoretically doubly and triply charged because positive charges can be attributed to the N terminus and to C-terminal lysine or arginine residues, respectively. Sialylated glycopeptides containing additional negative charges are therefore eluted prior to non-glycopeptides using strong cation exchange material. A similar principle has also been applied for the isolation of phosphopeptides by strong cation exchange chromatography (15) in the past.
In this work we devised a novel strategy by combining for the first time principles of aqueous two-phase partitioning and strong cation exchange chromatography for enhanced N-glycosylation site analysis using strong cation exchange enrichment (ENSAS). Applied to human platelets ENSAS led to the identification of 148 individual glycosylation sites on 79 proteins. Thereof 102 sites (69%) were not previously described in the Swiss-Prot database, and a high proportion of 75% plasma membrane-associated proteins was observed, e.g. low abundance G-protein-coupled receptors. Furthermore several new platelet proteins were found in this study with implications on current platelet research and potential clinical relevance.

EXPERIMENTAL PROCEDURES
Materials-All chromatographic chemicals used in this study were obtained from Sigma as analytical or higher grade. PNGaseF (Flavobacterium meningosepticum) and neuraminidase (Clostridium perfringens) were from Roche Applied Science. Sequencing grade trypsin was purchased from Promega, Madison, WI.
Platelet Purification-Human platelets were prepared by differential centrifugation and washing steps with minor modifications as described previously (6,16). Briefly fresh apheresis-derived and already leukocyte-depleted platelets (Department of Transfusion Medicine, University Wü rzburg, Germany) were centrifuged twice at 310 ϫ g for 15 min to remove remaining leukocytes or erythrocytes. The plateletcontaining supernatant was centrifuged at 380 ϫ g for 20 min at room temperature, and pelleted platelets were washed twice with 10 mM citric acid buffer containing 5 mM KCl, 145 mM NaCl, 14 mM glucose, and 1 mM MgCl 2 , pH 6.4, to remove residual plasma proteins from the apheresis preparation. The obtained platelet pellets were frozen in liquid nitrogen until further use.
Aqueous Two-phase Partitioning-For the enrichment of platelet plasma membranes two-phase partitioning as a variation of previously published methods was used (11). A 200-ml two-phase system consisting of 6.3% PEG 3350 and dextran T500 each in 15 mM Tris, pH 7.8, was prepared and equilibrated overnight at 6°C. Furthermore a 20-ml system (1:1 (v/v) of the equilibrated PEG and dextran phases) was used for separation of 100 mg (wet weight) of platelets. Lysis was performed by repetitive ultrasonic bursts, and subsequent phase separation was achieved by centrifugation at 500 ϫ g for 10 min at 6°C. The top PEG phase was removed and mixed with a fresh dextran phase followed by gentle mixing. This exchange was repeated once, and the final upper PEG phase was diluted 1:1 with water before membranes were pelleted by ultracentrifugation at 100,000 ϫ g for 1 h at 4°C in a TLA 100.4 rotor (Beckman Coulter, Krefeld, Germany). The resulting four pellets were subjected to 2-fold FIG. 1. Scheme of ENSAS enrichment. Platelet plasma membranes are enriched after cell lysis by aqueous twophase partitioning (steps 1 and 2) in a 6.3% polyethylene glycol/dextran polymer system. Upon isolation of membranes from the (upper) polyethylene phase and removal of cytosolic compounds by carbonate extraction (step 3), generated tryptic peptides (step 4) are subjected to strong cation exchange enrichment (step 5). Because of their low net charge, caused by additional sialic acid residues at the glycan side chains, glycopeptides are not retarded and can be collected in primary fractions. Enzymatic deglycosylation and simultaneous deamidation of asparagine to aspartic acid within NX(S/T) consensus sequences (step 6) permits later identification of the glycosylation site and the respective protein by mass spectrometry.
carbonate extractions in a volume of 3 ml of 100 mM sodium carbonate, pH 11.5, each as described elsewhere (45). A single pellet (or an equivalent amount of intact platelets) was solubilized in 200 l of 2.5% SDS, 25 mM NH 4 HCO 3 , pH 7.8, and 1,4-dithiothreitol was added to a concentration of 5 mM prior to incubation at 57°C for 15 min. Upon addition of iodoacetamide to a final concentration of 20 mM and incubation at 21°C for an additional 15 min, 1800 l of ethanol was added, and proteins were precipitated at Ϫ30°C for 4 h for removal of SDS. Proteolytic digestion of solubilized proteins in 400 l of NH 4 HCO 3 with trypsin (sequencing grade, Promega) was performed at 37°C overnight. For removal of undigested proteins, samples were ultrafiltrated with a-10 kDa molecular mass cutoff (Amicon Ultrafree-MC, Millipore, Schwalbach, Germany).
SCX Prefractionation-For HPLC enrichment of glycopeptide-containing fractions, a 2.1-mm-inner diameter ϫ 15-cm long column (PolySULFOETHYL Aspartamide, 200-Å pore size, 5-m particle size; Chromatographic Technologies, Basel, Switzerland) in combination with a Famos TM /Ultimate TM HPLC system (Dionex, Idstein, Germany) was used. A binary buffer system consisting of 5 mM NaH 2 PO 4 , pH 2.7 (buffer A) and 5 mM NaH 2 PO 4 , 15% acetonitrile, 500 mM NaCl, pH 2.7 (buffer B) was used to equilibrate the column at 1% B and a flow rate of 120 l prior to separation. Upon injection of 100-l sample aliquots, the column was washed for 7 min at 1% B followed by a gradual increase to 30% B (28 min) and 99% B (29 min). After rinsing the column for 7 min, the system was equilibrated to 1% B again. Fractions were collected in 60-s intervals by a ProteineerFC fraction collector (Bruker Daltonics, Bremen, Germany).
Self-made spin columns were constructed using C 18 pipette tips (Omix minibed; Varian, Darmstadt, Germany) packed with a 5-l volume of SCX material (see "SCX Prefractionation") that were placed within HPLC borosilicate sampling vials (CS-Chromatographie, Langerwehe, Germany). Spin columns were washed three times with 30 l of 5 mM NaH 2 PO 4 , 50% acetonitrile, pH 2.7, at 500 ϫ g. Samples were freeze-dried, solubilized in 30 l of washing buffer, and added on top of the columns; upon slow centrifugation at 500 ϫ g the flow-through was collected, and columns were subsequently discarded to avoid cross-contamination with further samples.
Glycosidase Treatment-Prior to SCX enrichment, 20-l peptide digest aliquots were diluted with 120 l of 50 mM ammonium acetate, pH 5.0, and treated with 0.08 unit of neuraminidase for 16 h at 37°C. To remove glycan side chains after SCX enrichment, digests with PNGaseF were performed at 5 units/100-l fraction size in 50 mM ammonium hydrogen carbonate, pH 7.8, for 16 h at 37°C.
Mass Spectrometric Analysis-Separation of peptide mixtures prior to mass spectrometric sequencing was achieved by nano-LC-MS/MS coupling. A Famos, Switchos TM , Ultimate nano-LC system (Dionex) was used to trap and desalt isolated peptide mixtures on a self-made 100-m-inner diameter ϫ 2-cm long precolumn (Ace C 18 , 5-m particle size, 10-Å pore size; HiChrom Ltd., Berkshire, UK) with 0.1% TFA as loading buffer. Separation on a self-made 75-m-inner diameter ϫ 150-mm long separation column (Ace C 18 , 3-m particle size, 100-Å pore size; HiChrom Ltd.) was performed at a flow rate of 270 nl/min and gradient slopes of 1% B/min and 0.5%B/min up to 55% B content, respectively. Solvent A was 0.1% formic acid in water, and solvent B was 0.1% formic acid in 84% acetonitrile.
A Qtrap 4000 linear ion trap mass spectrometer was used in the positive ion mode comprising 1) an enhanced multiple charge scan (380 -1500 amu, three spectra summed at 4000 amu/s) as survey scan followed by 2) enhanced resolution scans of selected precursors (single spectra at 250 amu/s) that were furthermore sequenced by 3) enhanced product ion scans (115-1500 amu, two spectra summed at 4000 amu/ s). Ion spray voltage was set to 2.3 kV, and only ions with charge states 2ϩ and 3ϩ were chosen for fragmentation. Dynamic exclusion time was set to 22 s after one occurrence of a respective target ion.
To further separate peptide mixtures, two independent procedures were added to the described analysis pathway: 1) on-line two-dimensional fractionation by SCX/reversed phase coupling and 2) gas phase fractionation in combination with 0.5% B/min reversed phase gradients. For SCX on-line fractionation a triphasic precolumn (100-m inner diameter, 1.5-cm C 18 /1.5-cm SCX/1.5-cm C 18 ) was used. Samples were applied in 0.1% formic acid and flushed onto the first C 18 phase. Upon elution to the SCX phase by a 4% B/min gradient, peptide subsets were stepwise eluted onto the second C 18 phase by injection of 1 l of 1, 5, 7.5, 10, 20, 30, 50, 70, 90, and 150 mM ammonium acetate, pH 2.7. Common reversed phase gradients as described above were used to separate individual subsets of peptides. Because the sample amount was not severely limited gas phase fractionation was applied by limiting the range of survey scans to smaller individual intervals (380 -500 m/z, 490 -600 m/z, 590 -700 m/z, 690 -900 m/z, and 890 -1500 m/z, all measured at 4000 amu/s with five spectra summed). Thereby the same sample was measured five times with triple play events as described above, fragmenting the three most intensive ions in the respective m/z intervals at a time.
Database Evaluation-Mass spectrometry-derived datasets were evaluated by database searches using the Mascot TM search algorithm (Version 2.1, Matrix Science, London, UK). Peak lists were generated from the raw data format using Analyst 1.4 software plugins (mascot.dll; Matrix Science/Applied Biosystems). All peaks with intensities below 0.1% of the base peak were omitted while data were centroided in the process. Mass deviance was set to 0.4 Da to readily identify asparagine to aspartic acid conversion of ϩ1 Da. A nonredundant human subset of the Swiss-Prot database (version from May 31, 2006; total 222,289 sequences, thereof 14,106 in the human subset; www.expasy.ch) was used for searches with trypsin specified as protease comprising one missed cleavage site. Spectra with a Mascot score Ͼ35 (significance threshold p Ͻ 0.05) and valid glycosylation consensus sequence were considered for further manual evaluation.
The Multi-Protein Survey System (Shanghai Center for Bioinformation Technology) was used for data evaluation and assignment of protein function (see Supplemental Table 1). For prediction of transmembrane domains TMHMM 2.0 was used (Center for Biological Sequence Analysis, Technical University of Denmark).

Enrichment of Plasma Membrane
Proteins-Purification of platelet plasma membranes by aqueous two-phase partitioning proved to be an efficient method for sample preparation prior to ENSAS enrichment. The complete procedure is readily performed in 4 -5 h. Preliminary results of one-dimensional SDS-PAGE coupled with nano-LC-MS/MS sequencing indicated over 200 different proteins with a high percentage of integral membrane proteins (data not shown). This finding also applies to proteins identified in the current study. Of 79 identified proteins (Table I), 89% could be assigned to membranes of organelles such as plasma membrane (64%), or membrane localization was predicted based upon primary sequence analysis but with unknown localization so far (14%) (Fig. 2). In addition, proteins that are present within both the plasma membrane and other organelles (Golgi, endosomes, etc.) could be identified (9%).
Because the complete strategy uses no gel-based separation, suitability for purification of membrane proteins is achieved; initially proteins are still embedded in biological membranes, and later on the detergent is removed directly before the proteolytic digest, which in turn produces smaller and thus more soluble peptides. The GRAVY index, an indicator for the hydrophobicity of identified protein sequences (see Supplemental Table 1), shows a broad scattering of values ranging from hydrophilic and abundant proteins such as fibrinogen (Ϫ0.78) over low abundance membrane receptors such as proteinase-activated receptor 4 (ϩ0.41) to extremely hydrophobic proteins such as the tetraspanin family with tetraspanin 13 (ϩ0.81). In general, numerous proteins with multiple membrane-spanning domains were detected; examples are the G-protein-coupled receptor PAR4 with seven predicted transmembrane domains and TM9S3 with nine or SID1 with 11 predicted membrane-spanning domains.
ENSAS Enrichment of Glycopeptides-In the current approach plasma membrane glycoproteins as potential initiators of vital platelet functions were targeted. Therefore, strong cation exchange chromatography was used for enrichment of sialylated glycopeptides. These peptides were collected in flow-through fractions with low contaminations by non-glycopeptides. Of 595 peptides identified by Mascot with a score above 35, 83% contained a modified asparagine within the N-glycosylation consensus sequence. Remaining peptides either had no modification at all or were modified at asparagines outside of the consensus sequences. However, the majority of these spectra were of inferior quality regarding noise levels and signal assignment by the algorithm. Several glycopeptides kept eluting after 10 min, although this observation was attributed to smearing effects because those peptides were identified in several adjacent fractions. At later gradient steps, Q08722

TABLE I Complete list of analyzed glycoproteins
Platelet glycoproteins were derived from aqueous two-phase partitioning and SCX prefractionation. For each protein the Swiss-Prot accession number followed by gene name and number of identified glycosylation sites are listed (for a complete list see Supplemental Table 1).
CD69 1 no distinct signals of glycopeptides could be identified probably due to quenching effects by the surpassing number of non-glycopeptides. To clarify the mechanism of sialoglycopeptide enrichment during SCX chromatography, a neuraminidase with broad specificity cleaving ␣(2-3)-, ␣(2-6)-, and ␣(2-8)-bound N-acetylneuraminic acid was used to remove potential negative charges from glycopeptides. This cleavage allows for answering whether the non-retention of glycopeptides is due to additional negative charges or to potential size exclusion effects. It was shown by Alvarez-Manilla et al. (17) that glycopeptides may also be enriched by size exclusion chromatography. Furthermore the SCX resin used in the current work has been demonstrated as a size exclusion stationary phase elsewhere (18). After the neuraminidase treatment, desialylated samples exhibited no significant enrichment of glycopeptides. Only six glycosylations sites could be assigned within the samples (Asn-478/P05106, Asn-37/P07359, Asn-198/P27701, Asn-344,Asn-356/P16284, Asn-374/P10909) including two sites for the high abundance proteins integrin ␤3 and platelet glycoprotein Ib ␣ chain. It was concluded that these identifications can be attributed to partially incomplete desialylation. Therefore, desialylated peptides seem to be as strongly retarded as their unglycosylated counterparts that elute later in the gradient. Furthermore size exclusion effects on ENSAS seem improbable due to the results of the neuraminidase treatment. Although the loss of sialic acids would reduce the size of glycopeptides, such a complete reduction of the enrichment effect seems improbable in respect to the remaining carbohydrate moieties. By targeting sialoglycopeptides of membrane glycoproteins, a total of 148 individual glycosylation sites distributed to 79 proteins could be identified (Table I). Upon comparing the sites with the Swiss-Prot database, 69% have not been determined previously (Fig. 2). Thereof 94 sites or 63% were only automatically annotated as potential glycosylation sites based on the presence of NX(S/T) consensus sequences. However, no experimental proof was provided in those cases so far. In comparison with former hydrazide and lectin affinity procedures of whole platelet lysates only a minor overlap of 16 sites (11%) was noted. Those sites contribute to the 46 hits (31%) termed as 'known' in the current study. Because the focus of ENSAS enrichment is set selectively to glycopeptides, it results in a dramatic reduction of sample complexity. Thereby the unfavorable balance of predominant identification of high abundance proteins in proteomics is shifted in favor of those with very low abundance. Of the 79 identified proteins, 46 were identified based on single glycopeptides, and a further 18 proteins were identified by two glycopeptides. Among the former were three G-protein-coupled receptors (PAR4, CCR4, and PI2R; see also Fig. 3 for spectrum of CCR4) with a few hundred copies per cell as well as a novel glycosylation site on a lesser known platelet receptor, G6b (19). Those sites could be found despite the high background of e.g. ITB3 with up to 80,000 copies per cell (20).
The reduction of complexity due to the focus on sialoglycopeptides will most often result in a loss of isoform resolution due to the reduced sequence coverage of the proteins. However, tumor necrosis factor receptor subfamily member 5 was identified by a single glycosylation site directly within an isoform-determining sequence. Although the membrane-bound isoform I contains an NKT consensus sequence within the peptide K2DLVVQQAGTNK2T, there is no such consensus motif at all in the soluble isoform II. Glycosylation site analysis may therefore help in individual cases to resolve such specific isoforms that may possibly not be detected in standard proteomics experiments.
The identification of N-glycosylation sites by mass spectrometry most often relies on the specific deamidation of asparagine to aspartic acid within the consensus sequence NX(S/T) (where X is not proline) upon cleavage of the glycan moiety by PNGaseF (21). In general, an artificial 1-Da mass shift from asparagine to aspartic acid as enzymatically introduced by the ENSAS procedure is not unique in proteins. Naturally occurring deamidation as well as preparation artifacts could induce the same shift. To assess the possibility of

FIG. 2. Aqueous two-phase partitioning and ENSAS confirm high number of potential N-glycosylation sites (A) with predominant assignment to platelet plasma membrane proteins (B).
A, distribution of N-glycosylation sites according to current Swiss-Prot annotation divided into known, "potential," "by similarity," and "unknown." The majority of sites were identified for the first time, although the proteins have been predicted to be glycosylated (potential, 63%). B, localization of identified proteins according to the Swiss-Prot database. Over two-thirds of hits were referred to as plasma membrane or co-localized. A further 14% of proteins were of unknown localization although predicted to be membrane-bound by TMHMM 2.0. false positive deamidation processes, SCX flow-through fractions were tested that were not PNGaseF-treated. However, few deamidated peptides could be detected. Therefore, the enrichment of sialylated glycopeptides seems to eliminate this major cause of false positives. Moreover reliable site determination was based upon three principles: 1) the peptide was readily identified by the Mascot algorithm, 2) the main Y ion series as well a2 ion signals were present as characteristic for the triple quadrupole linear ion trap mass spectrometer, and 3) the asparagine to aspartic acid shift was assigned to an NX(S/T) consensus sequence. In Fig. 3 two representative enhanced fragmentation spectra of CCR4 and TRPC6 are presented. The spectra exhibit all mentioned criteria, and accompanied by the fact that almost no non-glycosylated precursor ions were detected within individual preparations, those criteria ensure correct assignment of glycosylation sites. Therefore, no further H 2 18 O labeling is necessary to introduce an artificial 3-Da mass shift, which cannot be observed in natural processes as suggested by other studies.
Alternate Modes of ENSAS Enrichment-Frequently limited sample amounts may not allow for large scale ENSAS enrichment of glycopeptides. Therefore, we devised a modified system for glycopeptide enrichment comprising small spin columns for minute sample amounts. Spin columns were constructed of commercially available C 18 pipette tips filled with 5 l of chromatographic SCX material. After equilibration and sample application, glycopeptides in the flow-through fractions were analyzed. Subsequent application to platelet membrane proteins yielded 40 identified glycosylation sites within two redundant samples (see Supplement Table 2). The reduced numbers of confirmed sites in comparison with HPLC-based ENSAS can be attributed to the limited capacity of spin columns. Exceeding this limit, a multitude of non-glycosylated peptides is detectable in the flow-through because they are no longer retarded by the saturated SCX phase. The spin column procedure completely omits the need for a preparative HPLC. Furthermore sample amounts and volumes are readily compatible e.g. with PAGE-separated and in-gel digested samples. Thereby ENSAS can be combined with other isoform-resolving proteomics separation systems.
A further observation during two-dimensional SCX/reversed phase separation of PNGaseF digests of ENSAS samples offers the perspective for analysis of individual glycopeptides in complex mixtures. Frequently glycopeptides are quenched in the presence of coeluting non-glycopeptides and are therefore not accessible to mass spectrometric analysis. In twodimensional separations, the SCX phase is operated under conditions similar to those in ENSAS. Again sialylated glycopeptides are not retarded or are only minimally retarded and directly trapped on the subsequent reversed phase material. They can be eluted independently of the SCX-retarded peptides by a common reversed phase gradient or during the first SCX elution step. Analysis of individual protein digests can therefore yield direct information on their glycopeptides, glycosylation sites, and attached glycans by carefully studying individual glycopeptide-derived spectra. However, this is limited to low complexity digests from a few proteins so far. DISCUSSION As an initial step of glycomics strategies, the elucidation of glycosylation sites has gained interest in current proteomics research, although this analysis offers only minor information about the glycan attached. In the current study, a more focused work aiming at the determination of N-glycosylation sites specifically on plasma membrane proteins was conducted on the model of human platelets. In this context, aqueous two-phase partitioning followed by high salt extraction leads to a specific enrichment of membrane proteins and moreover to removal of high amounts of cytoskeletal proteins. The subsequently applied ENSAS workflow offers several advantages. 1) Consisting only of liquid chromatography steps, it is well suited for the analysis of hydrophobic membrane proteins.
2) The charge-dependent enrichment of glycopeptides results in high quality preparations of sialoglycopeptides with low background of other peptides. 3) Because unglycosylated background peptides are removed in the process, the complexity of the sample mixture is remarkably reduced, and FIG. 3. Enhanced fragment ion scans for identification of glycosylation sites. Examples of fragment ion scans for tryptic peptides of short transient receptor potential 6 (upper panel) and C-C chemokine receptor type 4 (lower panel). Annotation was performed for the main Y ion series as well as for internal fragmentations, immonium ions, and characteristic a2 ions; the modified asparagine (ϩ1 Da) is marked by an asterisk within the sequence. cps, counts/s. quenching effects by high abundance peptides during mass spectrometry can be avoided. 4) Miniaturization of the ENSAS process can be utilized to interface with common proteomics separation techniques. 5) As one of the upcoming methods in glycomics and proteomics, relative quantification of samples can be used for comparison of glycosylation events, e.g. during glycosylation disorders (22) or in bleeding phenotypes. Thereby relative quantification is supposed to produce results of clinical relevance by mapping glycosylation patterns and investigating new biomarkers as potential drug targets in platelets. In analogy to experiments reported by Kaji et al. (3) we used H 2 16 O/H 2 18 O endoglycosidase digests to label two identical preparations of ENSAS samples in the final deglycosylation step, and this yielded a 1:1 ratio of peak areas as demonstrated by the PECAM-1 peptide K2VLEN*STK2N (where N* indicates the modified asparagine) in Fig. 4. Therefore, relative quantification of samples seems to be feasible, although problems like overlapping isotopic peaks due to the inferior 2-Da mass difference and the rather late labeling step in the ENSAS process have to be solved. The former may be solved by bioinformatics (23) or by increasing the mass difference, e.g. using H 2 16 O/H 2 18 O-based tryptic digests (24). The late labeling step may nevertheless result in artificial differences of the quantification process. Ideally labeling should take place during the first steps of the procedure or up front as shown e.g. in SILAC (stable isotope labeling by amino acids in cell culture) procedures (25) which moreover do not change the charge distribution of the peptides.
In contrast to a similar study by Zhang et al. (4), using hydrazide enrichment after oxidation of carbohydrate moieties, ENSAS offers a faster enrichment of glycopeptides without the need for chemical derivatization, prolonged enrichment procedures, and sample cleanup steps. Furthermore the relative quantities of membrane proteins after ENSAS are higher in relative and absolute comparison; however, different samples were used in the studies (LNCaP prostate cancer epithelial cell line and human platelets in the current study, respectively). Nevertheless the presented method is limited to the analysis of sialic acid-containing glycopeptides. The SCX prefractionation step does not allow for analysis of glycopeptides with neutral carbohydrate side chains because these peptides are retarded on the chromatographic resin and are not separated from the bulk of non-glycopeptides. Therefore, ENSAS is suitable for the characterization of a glycopeptide subset, although non-sialylated glycopeptides may not be accessible via this method. However, non-glycosylated peptides also are copurified by the ENSAS process. Purification of phosphopeptides was shown by Beausoleil et al. (15) with a similar system from whole cell lysates. Because no phosphatase inhibitors were used in the current procedure, few phosphopeptides were detected in the SCX flow-through. However, they may be completely removed by an additional phosphatase treatment prior to the SCX enrichment. Furthermore N-terminally acetylated or C-terminal peptides of proteins were observed in the glycopeptide-containing fractions. Due to the reduction of net charges by acetylation or the lack of positive charge in the form of lysine/arginine these peptides may be poorly retarded by the SCX phase. These peptides were seldom identified and overall did not interfere with the detection of glycopeptides.
Besides mere analytical implications of the presented method, the high number of 102 newly confirmed glycosylation sites might stimulate reinvestigation of the functional aspect of platelet glycosylation events. Proteins like the integrin family or intercellular adhesion molecules are vital for platelet function, and their glycosylation status can influence platelet adhesion as shown previously (26). In contrast to our former study on whole platelets (6), the current protein population comprises subclasses not shown before. The tetraspanins, of which five members (tetraspanins 9, 13, 15, and 27 and CD63) could be identified in this study, share the structural detail of four transmembrane-spanning regions and are supposed to be involved in organizing multimolecular complexes in the plasma membrane (27,28 (29,30). Although similar in structure, each tetraspanin seems to specifically interact with a number of target molecules. In this context, CD63 has been shown to bind to the ␤ subunit of Na ϩ /K ϩ -transporting ATPase leading to its internalization in different cell types (31). This ATPase also was identified during our experiments by confirming a potential glycosylation site at Asn-124. In turn, Na,K-ATPase has been shown to interact with another protein derived from this study termed TRPC6. TRPC6, short transient receptor potential channel 6, is a non-selective cation channel that is supposed to be the store-independent calcium channel in platelets (32) (for the spectrum of TRPC6 peptide see Fig. 3). In addition two potential glycosylation sites on the well known purinoceptor P2X1 at Asn-184 and Asn-300 could be confirmed. Upon binding to ATP, P2X1 is responsible for fast Ca 2ϩ influx into platelets and functions thus as gated calcium channel (33,34). The field of calcium signaling has recently found renewed interest in the platelet field by the identification of stromal interaction molecule 1 (STIM1) as the potential Ca 2ϩ sensor in the endoplasmic reticulum leading to activation of calcium release-activated calcium channels by an as yet not fully known mechanism (35)(36)(37)(38). However, the function of several other identified proteins is yet unknown. In this context, transmembrane protein 16F (TM16F) and transmembrane 9 superfamily protein number 3 (TM9S3) can be stated as examples. Both proteins are predicted to contain a high number of transmembrane domains with seven to eight domains predicted for TM16F and nine domains predicted for TM9S3. Although the latter belongs to the nonaspanin family of proteins, TM16F has no such corresponding background.
Regarding the class of identified receptors, G6b is a putative member of the immunoglobulin class of cell surface receptors with a completely unknown function in platelets. Another isoform (G6f) was found in a former study on platelet glycosylation in our laboratory and was subsequently the subject of closer examination by other groups (39). Nevertheless G6b has so far only been studied in cell lines where it was found that the receptor binds heparin (40) and interacts with the tyrosine phosphatases SHP-1 and SHP-2 (41). Although it was detected several years ago by a proteomics study using diagonal chromatography of platelet N-terminal peptides (19), the current study not only elucidated an N-glycosylation site of G6b but moreover confirmed it as promising target for platelet research of plasma membrane proteins.
The detection of certain proteins such as multimerin, however, suggests a minor but expected contamination of the plasma membrane fraction by intracellular vesicles such as ␣-granules and storage vesicles. As a prominent example, multimerin 1 is a massive glycoprotein (42,43) with 23 potential glycosylation sites of which eight sites could be confirmed in this study. It functions as a storage protein for coagulation factor V until the activation of factor V leads to dissociation of the complex, thereby demonstrating a role of multimerin 1 in delivering and localizing factor V onto platelets prior to prothrombinase assembly (44). In general, many of the identified glycoproteins have not been put under detailed investigation to determine whether or not their glycosylation state influences e.g. function or ligand binding properties. Therefore, the current results can trigger further single protein analysis on the basis of this large scale in vivo glycosylation study. In this context, a major advantage of the ENSAS workflow is shown by the analysis of receptors such as PAR4 or CCR4 directly from platelet samples. Because platelets cannot be cultivated in vitro and their production by megakaryocytes is hampered by many obstacles, native platelet samples are the only option for protein analysis. Although in vivo glycosylations sites of low abundance cellular components are of special interest, they can hardly be studied using model cell lines in the case of platelets. We are therefore confident that the developed methods as well as presented results will be of further use in the field.