The Use of Multidimensional Liquid-Phase Separations and Mass

DSPAα1 is a single-chain plasminogen activator derived from Desmodus rotundus (vampire bat) salivary glands and is a serine protease that plays a rol...
1 downloads 0 Views 353KB Size
Download PDF
Anal. Chem. 1998, 70, 3572-3578

The Use of Multidimensional Liquid-Phase Separations and Mass Spectrometry for the Detailed Characterization of Posttranslational Modifications in Glycoproteins S. Udiavar,† A. Apffel,† J. Chakel,† S. Swedberg,† W. S. Hancock,*,† and E. Pungor, Jr.‡

Analytical/Medical Laboratory, Hewlett-Packard Laboratories, 3500 Deer Creek Road, Palo Alto, California 94304, and Berlex Biosciences, Richmond, California 94804

The goal of characterization of the proteome, while challenging in itself, is further complicated by the microheterogeneity introduced by posttranslational modifications such as glycosylation. A combination of liquid chromatography (LC), capillary electrophoresis (CE), and mass spectrometry (MS) offers the advantages of unique selectivity and high efficiency of the separation methods combined with the mass specificity and sensitivity of MS. In the current work, the combination of liquid-phase separations and mass spectrometry is demonstrated through the on-line coupling of electrospray ionization mass spectrometry (ESI-MS) and off-line coupling with matrix-assisted laser desorption/ionization time-of-flight mass spectrometry (MALDI TOF-MS). LC/ESI-MS yields real-time results while maintaining the separation obtained from the LC analysis. CE/MALDI TOF-MS offers high-mass detection and extremely low detection limits. The unique separation selectivity of CE relative to reversedphase HPLC separations of the members of a glycopeptide family was used to develop an integrated multidimensional analysis achieved by the off-line coupling of LC, CE, and MALDI TOF-MS. To demonstrate the applicability of these techniques to the characterization of the heterogeneity of posttranslational modifications present in glycoproteins, we will report on the study of the glycoforms present in a N-linked site in a single-chain plasminogen activator (DSPAr1). An important goal of the elucidation of the human genome is to probe the resulting massive amount of DNA sequence information for regions that code for novel protein sequences. After transfection of the novel DNA sequences into production cells, it will be possible to produce sufficient amounts of these proteins for the discovery of novel biological functions; see ref 1 as an example. A key challenge in such studies will be to appreciate the role of posttranslational modifications (PTMs) in the activity paradigm. It has been claimed that up to 80% of mammalian cells contain PTMs.2 Glycobiologists have repeatedly demonstrated †

Hewlett-Packard Laboratories. Berlex Biosciences. (1) Witt, W.; Maass, B.; Baldus, B.; Hildebrand, M.; Donner, P.; Scheuning, W. D. Circulation 1994, 90, 421. ‡

3572 Analytical Chemistry, Vol. 70, No. 17, September 1, 1998

the relationship between biological activity and carbohydrate structure.2,3 The activity of a protein in vivo is often a complex interplay between obvious effects of PTMs on activity to more subtle effects such as the influence of such modifications on the rate of clearance of a circulating protein, the specificity of the interaction of the protein with different receptors, and the targeting of the protein to specific subcellular organelles. In such a complex situation it is reasonable to expect that a complete description of the heterogeneity and distribution of PTMs in a given protein will be an important part of understanding the proteome and the subsequent discovery of novel biological pathways. One of the most common examples of PTMs is glycosylation, but unfortunately it is well-known that glycoproteins are among the most analytically challenging classes of biological molecules.4-7 For example, it has been estimated that recombinant tissue plasminogen activator (rtPA) could contain as many as 11 500 different glycoforms.8,9 Other examples of complex recombinant DNA-derived proteins include human erythropoietin (rHu-Epo)10 and blood coagulation factor factor VIIa.3 To characterize such proteins, it is necessary to examine the distribution of carbohydrate chains in the polypeptide core and to identify and quantify the glycoforms at each site. Without improved analytical methods one can see that such a study would require a huge amount of material as well as unacceptable amounts of scientific resources, particularly if the human genome studies result in a massive number of protein products. Clearly, a one-dimensional separation is not sufficient to characterize such complex proteins. In these cases, a multidimensional approach should be considered which uses different analytical methods that analyze the sample from independent directions. Many methods have been studied to characterize (2) Kornfield, R.; Kornfield, S. Annu. Rev. Biochem. 1985, 54, 631-664. (3) Liu, D. T. Trends Biotechnol. 1992, 10, 114-119. (4) Rudd, P. M.; Scragg, I. G.; Coghill, E.; Dwek, R. A. Glycoconjugate J. 1992, 9, 86-91. (5) Harvey. D. J. J. Chromatogr., A 1996, 720, 429-446. (6) Lo-Guidice, J. M.; Lhermitte, M. Biochem. Chromatogr. 1996, 10, 290296. (7) Kakehi, K.; Honda, S. J. Chromatogr., A 1996, 720, 377-393. (8) Spellman, M. W. Anal. Chem. 1990, 62, 1714-1722. (9) Guzzetta, A. W.; Basa, L. J.; Hancock, W. S.; Keyt, B. A.; Bennett, W. F. Anal. Chem. 1993, 65, 2953-2962. (10) Rush, R. S.; Derby, P. L.; Strickland, T. W.; Rohde, M. F. Anal. Chem. 1993, 65, 1834-1842. S0003-2700(98)00405-3 CCC: $15.00

© 1998 American Chemical Society Published on Web 07/31/1998

glycoproteins,4,6,11-14 but peptide mapping has been one of the most widely tools.10,11,13,15 In this application, an important advantage of capillary electrophoresis (CE) is its complimentary nature with reversed-phase HPLC (RPLC).16,17 In RPLC, the separation is largely determined by the hydrophobicity of the peptide moiety of the glycopeptide, and thus these peptides coelute in an unpredictable manner, with the much more abundant nonglycosylated peptides often hidden in crowded regions of the map. In such a situation, even on-line electrospray ionization mass spectrometry (ESI-MS) will be able to detect only the major glycoforms present in a sample and the limit of detection of minor components will be a few percent of the larger peaks in the map.9,15 Whereas, in HPCE the mobility of the sample is dependent on its charge-to-mass ratio,16 and in the analysis of glycopeptides by capillary zone electrophoresis, it is possible to separate differences in both charged (sialic acid, due to charged differences at pH >3.0) and noncharged moieties (e.g., mannose), due to differences in hydrodynamic volume. Thus, previous studies have shown it is possible to optimize the CE separation such that the glycopeptides migrate largely separated from the other peptides in the map.15,18 It has been shown that electrospray ionization mass spectrometry used in combination with RPLC can enable the characterization of peptide maps and CID data can be used to detect the elution position of glycopeptides.13 Matrix-assisted laser desorption/ionization time-of-flight MS (MALDI TOF-MS) has recently been shown as a powerful new method that can be ideally suited for the rapid and sensitive analysis of large glycoproteins even in the presence of substantial microheterogeneity.18 Useful spectra have been reported for samples such as the plasminogen activator DSPA,15 a monoclonal antibody,19 interleukin 4 receptor.20 Also, MALDI TOF-MS has been used to analyze an unfractionated mixture of peptides present in the digest of DSPAR1.15 Again, as in the case of ESI-MS, not all of the peptides could be observed by such a process and the heterogeneity of glycosylation resulted in poor delectability of these fragments. However, the performance of MALDI TOF-MS was substantially improved by prefractionation of pools of the peptides by either RPLC or CE.18 It was the goal of this study, therefore, to use MALDI TOF-MS to explore the heterogeneity present at an N-linked site of the glycoprotein DSPA after the combined fractionation of RPLC and CE. (11) Kelly, F.; Locke, S. J.; Ramaley, L.; Thibault, P. J. Chromatogr., A 1996, 720, 409-427. (12) Taverna, M.; Baillet, M.; Schlu ¨ er, M.; Baylocq-Ferrier, D. Biochem. Chromatogr. 1995, 9, 59-67. (13) Huddleston, M. J.; Bean, M. F.; Carr, S. A. Anal. Chem. 1993, 65, 877884. (14) Yang, Y.; Orlando, R. Rapid Commun. Mass Spectrom. 1996, 10, 932936. (15) Apffel, A.; Chakel, J.; Udiavar, S.; Hancock, W. S.; Souders, C.; Pungor, E., Jr. J. Chromatogr., A 1995, 717, 41-60. (16) Grossman, D.; Colburn, J. C.; Lauer, H. H.; Nielson, R. M.; Riggin, R. M.; Sittampalam, G. S.; Rickard, E. C. Anal. Chem. 1989, 61, 1186-1191. (17) Apffel, A.; Chakel, J.; Hancock, W. S.; Souders, C.; Timkulu, T.; Pungor, E., Jr. J. Chromatogr., A 1996, 732, 27-42. (18) Chakel, A.; Pungor, E., Jr.; Hancock, W. S.; Swedberg, S. A. J. Chromatogr., B 1997, 689, 215-220. (19) Ashton, D. S.; Beddell, C. R.; Copper, D. J.; Craig, S. J.; Lines, A. C.; Oliver, R. W. A.; Rajan, N.; Tsarbopoulos, A.; Kumarasamy, R.; O’Donnell, R.; Taremi, S. S.; Baldwin, S. W.; Seelig, G. F.; Fan, X.; Pramanik, B.; Le, H. V. Biochem. Biophys. Res. Commun. 1995, 206, 694. (20) Smith, M. A. Anal. Chem. 1995, 67, 835-842.

DSPAR1 is a single-chain plasminogen activator derived from Desmodus rotundus (vampire bat) salivary glands and is a serine protease that plays a role in clot lysis.1 The recombinant product, expressed in Chinese hamster ovary cells, consists of a 441-amino acid sequence with an average (nonglycosylated) molecular mass of 49 508. From initial carbohydrate studies, DSPAR1 was shown to consist of four O-linked and two N-linked sites of glycosylation.15,21 In the current study, we chose to analyze the glycoforms of fragment K21 (see later) as a typical example of a highly heterogeneous N-linked oligosaccharide. In this paper, the endoproteinase Lys-C digest of DSPAR1 was initially analyzed by liquid chromatography (LC)/ESI-MS, CE, and MALDI TOF-MS. Fractions collected from the LC separation were further resolved by CE and then analyzed by MALDI TOFMS. The results of this study, which involves triple hyphenation of separation methods, can allow the comparison and then integration of different degrees of orthogonal separations needed for the accurate identification of the large number of glycoforms typically found in a heterogeneous glycoprotein. EXPERIMENTAL SECTION 1. RPLC. The HPLC separation was performed on a HewlettPackard 1090 liquid chromatography system with DR5 ternary solvent delivery system, diode-array UV/visible detector (DAD), autosampler, and heated column compartment (Hewlett-Packard Co.). All HPLC separations were done using a YMC ODS-AQ 3-mm particle, 120-Å pore size reversed phase column. A standard solvent system of H2O (solvent A) and acetonitrile (solvent B), both with 0.09% TFA, was used with a flow rate of 0.2 µL/min. The gradient for the separation was constructed as 0-60% B in 60 min. The column temperature was maintained at 45 °C throughout the separation. 2. LC/ESI-MS. Mass spectrometry was done on a HewlettPackard 5989B quadrupole mass spectrometer equipped with extended mass range, high-energy dynode detector (HED), and a Hewlett-Packard 59987A API-electrospray source with high-flow nebulizer option. Both the HPLC and MS were controlled by the HP Chemstation software allowing simultaneous instrument control, data acquisition, and data analysis. The high-flow nebulizer was operated standard with N2 as nebulizing (1.5 L/min) and drying (15 L/min at 300 °C) gases. A stepped fragmenter voltage was used to generate collisionally induced dissociation (CID) glycomarkers at m/z 204 (HexNAc), 292 (NANA), and 366 (HexNAc + Hex) To counteract the signal-suppressing effects of trifluroacetic acid on electrospray LC/MS, a previously reported method,22 referred to as the “TFA Fix” was employed. The TFA Fix consisted of postcolumn addition of 75% propionic acid, 25% 2-propanol at a flow rate of 100 µL/min. The TFA Fix was delivered using a HP 1050 HPLC pump and was mixed with a zero dead volume tee into the column effluent after the DAD detector and after the column-switching valve. Column effluent was diverted from the MS for the first 10 min of the chromatogram, during which time excess reagents and unretained components eluted. (21) Muschick, P.; Zeggert, D.; Donner, P.; Witt, W. Fibrinolysis 1991, 7, 284290. (22) Apffel, A.; Fischer, S.; Goldberg, G.; Goodley, P. C.; Kuhlmann, F. E. J. Chromatogr., A 1995, 712, 177-190.

Analytical Chemistry, Vol. 70, No. 17, September 1, 1998

3573

For peptide mapping, MS data were acquired in scan mode, scanning from 200 to 1800 Da at an acquisition rate of 1.35 Hz at a 0.15-Da step size. Unit resolution was maintained for all experiments. Data were filtered in the mass domain with a 0.03Da Gaussian mass filter and in the time domain with a 0.05-min Gaussian time filter. For the in-source collisionally induced dissociation (CID) method for detecting fragments indicative of glycopeptides, the CapEx voltage was dynamically set to 300 V for m/z 200-375 and 100 V for m/z 400-2000. 3. CE. CE was performed using the HP3DCE system with diode-array detection and Chemstation software for computer control and data acquisition/analysis. For CE fraction collection work, bovine serum albumin (BSA)-treated23 fused-silica capillaries (75 mm i.d., Leff 56 cm, -500 V/cm, 28 °C, ∼49 µΑ) were used with 25 mM ammonium phosphate buffer, pH 3.00. Fraction collection was performed by controlled application of pressure [5000 Pa (50 mbar)] for zone elution. The receiving vials for fraction collection contained a minimal volume of water (∼5 µL), sufficient to wet the end of the capillary and to avoid excessive dilution. 4. MALDI TOF-MS. Matrix-assisted laser desorption timeof-flight MS was performed on a HP G2030A system. All spectra shown were obtained in the positive ion mode. The predominant matrixes used were R-cyano-4-hydroxycinnamic acid (RCN) and 2,6-dihydroxyacetophenone (DHAP) mixed with a diammonium hydrogen citrate (DAHC) coadditive. The RCN matrix was prepared as a standard solution (from doubly recrystallized and ion-exchanged RCN) in 2:1 H2O with 1.3% TFA/acetonitrile. The DHAP/DAHC matrix was prepared with 30 mg of DHAP and 44 mg of DAHC in 1 mL of 1:1 H2O/acetonitrile. With the DHAP/ DAHC matrix, better results were often obtained by addition of 1 µL of 2:1 H2O with 0.10% TFA/acetonitrile. Typically, for the collected CE fractions where the final sample volume is ∼6 µL, the sample was deposited in two different positions, one for each matrix (3 × 1 µL with vacuum-drying in the HP 2024A sample preparation accessory). A 1-µL sample of each matrix was then applied, followed by rapid vacuum evaporation, and subsequently put into the TOF mass spectrometer for analysis. Typically, 50100 laser shots were summed for each spectrum. Mass calibration was internal, using added h-angiotensin I (1296.5 Da) and h-insulin (5807.7 Da). 5. Chemicals and Reagents. HPLC-grade acetonitrile and trifluoroacetic acid (TFA), as well as EDTA were obtained from J. T. Baker (Phillipsburg, NJ). Distilled, deionized Milli-Q water (Millipore, Bedford, MA) was used. Urea was obtained from Gibco (Gaithersburg, MD), DL-dithiothreitol (DTT), iodoacetic acid, bicine, NaOH, and CaCl2 were obtained from Sigma (St. Louis, MO). The enzyme endoproteinase Lys-C (Wako BioProducts, Richmond, VA) was used for mapping. Recombinant DSPAR1 (from Berlex Bioscience, Richmond, CA) was purified from CHO cells, and the starting concentration was 0.5 mg/mL. Buffer components were purchased from Sigma and were of the highest purity available. 6. Sample Preparation. The endoproteinase Lys-C digest of DSPAR1 was prepared as follows. DSPAR1 (5 nmol) was desalted using a 10K NMWL centrifugal ultrafiltration tube at (23) Swedberg, S.; Herold, M. Poster presented at the 7th Symposium of the Protein Society, 1993.

3574 Analytical Chemistry, Vol. 70, No. 17, September 1, 1998

3000g for 20 min. The sample was denatured in 6 M guanidine hydrochloride. The sample was then reduced by addition of 100 µL (9 mg/mL) of dithiothreitol in 100 mM ammonium bicarbonate at pH 7.3 and incubated at 37 °C for 30 min. Alkylation of the reduced protein was done by a further 30-min incubation with 100 µL of iodoacetic acid (26 mg/mL) in ammonium bicarbonate buffer. The reduced and alkylated DSPAR1 was then desalted using the 10K NMWL centrifugal ultrafiltration unit (Millipore Corp.) for 1 h at 3000g. The reduced, alkylated, desalted DSPAR1 was then digested with endoproteinase Lys-C (Promega Corp) in 100 mM ammonium bicarbonate, pH 7.2, at 37 °C with an enzymeto-substrate (mass) ratio of 1:20 for 18 h. The resulting digestion was at a final concentration of 8 pmol/µL. The final digests were acidified and stored at 5 °C until use. Injections (250 µL) were made at the level of 2 nmol/250 µL. RESULTS AND DISCUSSION 1. Analytical Strategy. DSPAR1 is a complex glycoprotein that has been studied under chromatographic, electrophoretic methods and by MALDI TOF-MS.15 However, these methods are not sufficient to give all the information required to define the structure of the posttransitional modifications present in this protein. Peptide mapping has been used successfully to further study structural characteristics of such proteins. In peptide mapping, proteases are used to digest the protein, and the resulting fragments are then separated by either RPLC or HPCE. Due to the heterogeneity of glycosylation, any fragments that contain these modifications will be present in the map in a variety of forms that will be partly resolved by the separation method. Previous studies have shown that no one separation method is adequate for the characterization of a glycoprotein digest.15 To plan a multidimensional approach to this complex challenge, we decided to consecutively analyze the mixture of peptides from DSPAR1 by RPLC, CE, and MALDI TOF-MS. In the case of RPLC, it is difficult to predict the elution point of glycopeptides, as retention time is largely determined by the hydrophobicity of peptides (see the shaded areas in Figure 1). Also, the intensity of the UV or on-line MS detector signal will be low due to the microheterogeneity caused by glycosylation. Previous studies15 have used a so-called “sugar” scan which can be generated by in-source collisionally induced dissociation (CID).13 On the basis of the CID data, one can see that the glycopeptides elute in crowded regions of the map and this demonstrated the need for additional fractionation methods. In previous studies on CE,24 we have reported on the use of capillaries that were coated with BSA. Such capillaries have been shown to be useful in bioanalysis because of the properties of BSA as a biofouling resistant surface, which reduces the proteinwall interaction, giving maximum recovery. We reported on the analysis of the peptide map of DSPAR1 on BSA-coated capillaries at pH 7.0 and demonstrated that many of the glycopeptides eluted late in the map as relatively broad peaks. For fused-silica capillaries, at the low pH, the electroosmotic flow (EOF) is normally cathodic. However, using dynamic deactivation of the capillary [for example, cetyltrimethlyammonium (24) Swedberg, S. Anal. Biochem. 1990, 185, 51-56.

Figure 2. Elution profile of the digest of DSPAR1 by RPLC with UV and mass spectrometric detection (TIC trace shown in upper profile). The elution positions of the N-linked glycopeptide families (K4 and K21) are detected by the glycomarkers profile (lower trace) which was determined by CID. The solid bar shows the fraction collected for further study. See the Experimental Section for the detailed explanation. Figure 1. Separation of the peptide map (Lys-C digest) of DSPAR1 by (a) RPLC, (b) CE, and (c) MALDI TOF-MS. The elution points of the glycopeptide families are shown by boxes. The separation conditions are described in the Experimental Section.

bromide (CTAB; Polybrene)], this flow can be reversed.25 Furthermore, others demonstrated that, using bonded amphoteric phases, the magnitude and direction of the EOF may be tuned as a function of pH.24 In the case of the amphoteric BSA phase, the direction of the EOF can be reversed at pH values less than 4.0.26 For example, with this capillary, at pH 3.00 using phosphate buffer, the EOF is appreciably anodic (-1.2 mm/s), which is opposite to that obtained on fused-silica capillary at pH 7.00 (1.00 mm/s), using a phosphate buffer. With the BSA capillary at pH 3.00, therefore, the elution order of the peptides can be reversed, and in the case of one of the major N-linked glycopeptide families, the flow reversal resulted in the early migration of the glycosylated peptides (see Figure 1B). These results obtained by CE can be contrasted with the complex behavior of glycopeptides in RPLC (Figure 1A), where identification of these species was hindered by coelution with other peptide fragments. In the third approach to analysis of the digest, we performed MALDI TOF-MS on the peptide mixture, without any preseparation. While this approach gives many of the masses, again the picture is incomplete because of the variable ionization of different peptides. Typically with MALDI TOF-MS, only 60-80% of the expected fragments can be observed. A typical result for DSPAR1 (25) Yao, Y. J.; Loh, K. C.; Chung, M. C. M.; Li, S. F. Y. Electrophoresis 1995, 16, 647-653. (26) Maa, S.; Hyver, K. J.; Swedberg, S. A. J. High Resolut. Chromatogr. 1991, 14, 65-67.

analysis is shown in Figure 1c and agrees with previous studies in that up to 30 of the expected 51 peptides can be seen.15 Along with these peptides, glycopeptides of the expected mass 25004000 kDa can be seen at low intensity (see highlighted areas in Figure 1c). Therefore, a fractionation step is needed before MALDI TOF-MS for more complete analysis of glycopeptides, although MALDI TOF-MS of the mixture can provide a useful, quick fingerprint of a protein digest. From these initial comparisons, we developed an approach to characterize carbohydrate heterogeneity in more detail by the following combination of separation methods. First the sample is fractionated by LC, since LC allows larger sample loads, followed by CE and then MALDI TOF-MS. We believed that this approach would give a more complete picture of a heterogeneous sample than the single high-resolution methods described above, because of the complementarity of an appropriate combination of three orthogonal methods. One problem with this approach was sample handling of a very small amount of the heterogeneous samples. We started with 10 nmol of the protein digest for separation by LC, but only 2 nL of the resulting LC fraction was injected into the CE apparatus. Approximately 50% of the fraction from the CE run was then analyzed by MALDI TOF-MS. If one assumes 100% recovery at the end of each step, then the maximum sample amount was approximately 4-5 fmol at this stage, although one could expect that due to microheterogeneity and sample losses that the actual amount was considerably less. Therefore, a goal of this study was to develop sample handling protocols compatible with such complex samples. 2. Peptide Map. RPLC/ESI-MS. The on-line combination of RPLC and electrospray ionization mass spectrometry was Analytical Chemistry, Vol. 70, No. 17, September 1, 1998

3575

Table 1. Partial Summary of Results (HPLC from Fraction 28) MALDI TOF-MS data (M + H +)

Figure 3. Separation of fraction 28 by CE (BSA capillary, 50 mM phosphate buffer, pH 3.00.) The results of subsequent MALDI TOFMS identification of glycoforms are shown in the captions.

consistent with the previous observation of two sites, namely, K4 and K21 of N-linked glycosylation.15,21 Figure 2 shows the elution profile from the RPLC/ESI-MS analysis. Part a shows the separation monitored by total ion current (TIC), and part b shows the CID trace where the elution position for glycopeptides is determined by the observation of diagnostic low-mass fragments indicative of saccharides. In the current study, we chose to further analyze the glycoforms of fragment K21, and based on the ESIMS/CID spectra, the main glycosylated peptide K21 was found to be in the fraction 28 (see solid bar), with small amounts in the previous fractions 27 and 29. CE. Fractions 27, 28, and 29 had been shown to contain glycosylated peptide K21 and were then further studied using CE with the BSA-coated capillaries; fraction 28 was shown to contain the majority of glycopeptides (data not shown). The result of the CE analysis of fraction 28 is shown in Figure 3. As can be seen from the figure, the fraction was indeed further resolved by CE, with the separation giving at least three peaks. Each major peak grouping of fraction 28 from the CE separation was then collected manually for further identification of the glycoforms by MALDI TOF-MS (see the Experimental Section for details). The successful purification of the components in fraction 28 was demonstrated by reinjecting the purified peak under similar CE conditions as shown in Figure 3 (data not shown). It was decided to use fraction 28 for the next step of the multidimensional analysis. MALDI TOF-MS. After the combined fractionation of the total digest by RPLC and CE, the MALDI TOF-MS analysis gave a much more accurate assessment of the glycopeptide heterogeneity in the sample. An important result from the MALDI TOF-MS analysis was that a variety of different carbohydrate structures could be assigned to each of the three peaks seen in the CE separation (see Figure 3 and Table 1). While the detailed separation mechanism is not known at this stage, it is clear that CE can separate the glycopeptides by both differences in charge and hydrodynamic volume (see later for the MALDI TOF-MS results). In this analysis, the intensity of the signal for the glycopeptides has been increased significantly over the result for the unfractionated digest, probably due to the removal of lowmolecular-weight components and nonglycosylated peptides. One of the advantages of MALDI TOF-MS is that different matrixes can be investigated to optimize the MS measurement. 3576 Analytical Chemistry, Vol. 70, No. 17, September 1, 1998

identified structures Lys-C fragment K21 +

2728.5 3019.6 3384.7 3749.8

CE Fraction 28A NANA1Hex5HexNAc4Fuc NANA2Hex5HexNAc4Fuc NANA2Hex6HexNAc5Fuc NANA2Hex7HexNAc6Fuc

2420.4 2711.5 3076.7 3441.9

CE Fraction 28B NANA1Hex4HexNAc4 NANA2Hex4HexNAc4 NANA2Hex5HexNAc5 NANA2Hex6HexNAc6

2725.6 2437.7 2640.8 2802.8 3005.9 3167.9

CE Fraction 28C Hex4HexNAc4Fuc Hex5HexNAc4Fuc Hex5HexNAc5Fuc Hex6HexNAc5Fuc Hex6HexNAc6Fuc Hex7HexNAc6Fuc

Figure 4. Comparison of fraction 28 by using two matrixes in MALDI TOF-MS analysis. Part a shows the result with RCN/TFA matrix, part b with DHAP/DAHC.

Two common problems with the analysis of glycopeptides by this technique are the loss of sialic acid during the mass measurement and the observation of significant Na+ and K+ adduction. In an attempt to reduce these artifacts, RCN and DHAP/ diammonium hydrogen citrate (DAHC) were compared as suitable

Figure 5. MALDI TOF-MS spectra of CE fraction 28A. Structure of the oligosaccharide that is consistent with the observed mass is shown directly above the corresponding peak. The observed mass differences (and most likely carbohydrate difference) are shown at the bottom of the figure.

Figure 6. MALDI TOF-MS spectra of CE fraction 28B. All conditions same as Figure 5.

matrixes. While this matrix yielded greater sensitivity for the lower mass range, we observed two significant disadvantages. First, we did observe minor loss of sialic acid in some samples and also the spectra often showed significant salt (Na+ and K+) adduction (see Figure 4a). As shown in the Figure 4b, the DHAP/ DAHC matrix provided softer ionization and thus less sugar fragmentation, as well as fewer salt adducts. In this study,

however, due to very low sample amounts, the key issue was sensitivity and thus the RCN/TFA matrix was used in the following analyses. Figures 5-7 and Table 1 show the mass spectra of the CE fractions from 28 A-C, respectively, using the RCN/TFA matrix. Table 1 summarize the results from MALDI TOF-MS on CE fraction 28, and analysis of the data indicated that all of the major components in the electopherogram had masses that were consistent with glycosylated variants of peptide K21. In Figure 5, the MALDI TOF-MS spectra showed two major peaks which can be compared with a similar number of peaks in the CE analysis of fraction 28A. These two major peaks with nonprotonated masses of 3019.6 and 3384.7 Da were separated from each other by m/z 365 (a difference that corresponds to the mass of Hex + HexNAc, which is commonly encountered in N-linked oligosaccharides). Previously we found that the composition of the carbohydrate fraction of a peak having the mass 3019.6 Da is NANA2Hex5HexNAc4Fuc (+K21, where Hex ) hexose, Fuc ) fucose) and the peak having the mass 3384.7 Da is NANA2Hex6HexNAc5Fuc (+K21).18 Along with these two major peaks, two minor peaks were also observed in MALDI TOF-MS spectra of fraction 28A (see Figure 5). These were separated from the major peaks by m/z 365 (-Hex + HexNAc) and m/z 292 (-sialic acid, or NANA), respectively. Sialic acid heterogeneity is commonly encountered in N-linked oligosaccharides. It is clearly seen from the data that in the carbohydrate linkages of the fraction 28A both sialic acid (NANA) and fucose (fuc) are present. Analytical Chemistry, Vol. 70, No. 17, September 1, 1998

3577

Figure 7. MALDI TOF-MS spectra of CE fraction 28C. All conditions same as Figure 5.

In the spectra of CE fraction 28B, one major peak with the mass 3076.7Da was observed with the predicted structure K21 + NANA2Hex5HexNAc5 and can be compared essentially to one peak in CE (see Figure 6). Along with the major peak, three minor peaks were also observed which were separated by m/z 365 (-Hex + HexNAc) from the major peak and m/z 292 (-NANA) from each other; see Figure 8b. No fucosylated structures were observed in fraction 28B . The MALDI-TOF spectra of HPCE fraction 28C had a more complex structure. There were five minor peaks in the MALDI TOF-MS spectra along with a major peak which had a mass 2802.8 Da, with the predicted structure K21 + Hex5HexNAc6Fuc. These peaks were separated from each other by m/z 161 (-Hex) and m/z 243 (-HexNAc), as shown in Figure 7. In the case of fraction 28C, no sialylated structures were observed. The above findings show that peak 28A had sialylated as well as fucosylated complex structures, peak 28B had a sialylated complex structures, whereas peak 28C had fucosylated complex structures that contained little sialic acid. It was noted that all the masses observed in the MALDI TOF-MS study were consistent with a family of glycopeptides related to the K21 fragment of DSPAR1. This was consistent with earlier RPLC/ESI-MS studies

3578 Analytical Chemistry, Vol. 70, No. 17, September 1, 1998

that showed the other glycosylated fragments eluted in different regions of the map. CONCLUSION In this study, we have shown that HPLC, CE, ESI-MS, and MALDI-TOF are highly complimentary techniques for examining glycoproteins. A single dimensional separation of either HPLC or CE does not provide sufficient separation to characterize all of the major glycoforms. Furthermore, combination of HPLC/ESIMS and LC/CE/MALDI TOF-MS simplifies the degree of complexity of individual samples to allow tentative identification of >30 glycoforms present at a single site in DSPAR1. While such analytical studies are instrument intensive, we believe that a more complete elucidation of the heterogeneity of posttranslational modifications is a key step in the elucidation of the proteome. Thus, future instrument developments such as hyphenation of these methods, on-line coupling of MALDI TOF-MS to separations, steps to ensure greater throughput of samples and better data integration will naturally follow. Received for review April 14, 1998. Accepted June 16, 1998. AC980405Q