Anal. Chem. 2004, 76, 3112-3118
Selective Detection of Glycopeptides on Ion Trap Mass Spectrometers Barbara Sullivan, Theresa A. Addona,* and Steven A. Carr*,†
Protein Science and Technology Department, Millennium Pharmaceuticals, 45 Sidney Street, Cambridge, Massachusetts 02139
Generation of carbohydrate-specific marker ions during LC-ESMS of digested glycoproteins has been demonstrated to be a highly selective and sensitive approach for detection of glycopeptides. In principle, any mass spectrometer can produce and selectively detect carbohydrate marker ions provided that the instrument is capable of collisional excitation in the region prior to the first mass analyzer sufficient to form abundant oxonium ions. This approach has yet to be demonstrated on 3D ion trap mass spectrometers, which have become widely used for proteomic applications. Here we report the successful development and optimization of carbohydrate marker ion detection on a LCQ Deca 3D ion trap utilizing this scan function. Human r-1 acid glycoprotein and a therapeutic monoclonal antibody were chosen to illustrate this methodology. Marker ion detection during LC-ESMS facilitated collection of glycopeptide-containing fractions. Analysis of the glycopeptides in these fractions by MS identified the specific glycosylation sites and enabled the prediction of the family of glycoforms at each attachment site. Using these optimized conditions, marker ion detection and glycopeptide analysis could be achieved with as little as 10 pmol of a glycoprotein. Glycosylation is one of the most common and relevant types of posttranslational modifications of proteins, affecting nearly half of all proteins in a cell with respect to protein structure, biological activity, and solubility. In general, there are two types of protein glycosylation. The first type consists of attachment of carbohydrate through the amino group of asparagine for an N-glycosidic linkage. N-Linked glycoproteins typically contain a core structure of two N-acetylglucosamines and three mannoses. Oligosaccharides attached to protein through N-glycosidic linkages can be quite complex, consisting of bi-, tri-, and tetraantennary structures.1,2 Most N-linked glycosylation occurs at consensus sequences of N-X-Ser/Thr, where X can be any amino acid except proline. The second type of glycosylation, O-linked glycosylation, consists of attachments to proteins by a covalent linkage between monosaccharide N-acetylgalactosamine and the hydroxyl group of serine or threonine.1,2 * Corresponding authors. E-mail:
[email protected]. † Present address: Broad Institute, 320 Charles St., Cambridge, MA 02139. (1) Dennis, J. W.; Granovsky, M.; Warren, C. E. BioEssays 1999, 21, 412421. (2) Kornfeld, R.; Kornfeld, S. Annu. Rev. Biochem. 1985, 54, 631-664.
3112 Analytical Chemistry, Vol. 76, No. 11, June 1, 2004
Complete structural characterization of the oligosaccharides attached to a glycoprotein can provide valuable information on the relationship between protein structure and function and can also be important for recombinant protein production. A variety of methods have been used to identify protein-bound carbohydrates such as NMR, lectin affinity chromatography, anionexchange chromatography, reversed-phase chromatography, and capillary electrophoresis.3-5 However, these approaches generally require released oligosaccharide, and therefore, heterogeneity of the glycoforms in the context of the amino acid sequence is lost. Determining which sites of a glycoprotein are glycosylated can be challenging due to the heterogeneity of the glycoforms, as well as the sensitivity required to detect low-abundance glycopeptides. Mass spectrometry (MS) has become an increasingly valuable analytical technique to applications in proteomics and in studies of posttranslational modifications. Mass shifts resulting from covalent modification of peptides or proteins can be detected, and specific sites of attachment can be identified with very high sensitivity.6-14 For glycosylation analysis, structural information and characterization of the attached oligosaccharides can often be obtained in the form of putative structures based upon their apparent mass. Several groups have reported MS techniques that are tailored to the analysis of posttranslationally modified proteins.15-23 Carr and co-workers have shown the utility of using (3) Schmid, K.; Nimberg, R. B.; Kimura, A.; Yamaguchi, H.; Binette, J. P. Biochim. Biophys. Acta 1997, 492, 291-302. (4) Sutton, C. W.; Poole, A. C.; Cottrell, J. S. Tech. Biochem. IV 1993, 62, 109116. (5) Dwek, R. A.; Edge, C. J.; Harvey, D. J.; Wormald, M. R. Annu. Rev. Biochem. 1993, 62, 65-100. (6) Watts, J. D.; Affolter, M.; Krebs, D. L.; Wange, R. L.; Samelson, L. E.; Aebersold, R. J. Bio. Chem. 1994, 269, 29520-29529. (7) Schindler, P. A.; Settineri, C. A.; Collet, X.; Fielding, C. J.; Burlingame, A. L. Protein Sci. 1995, 4, 791-803. (8) Carr, S. A.; Huddleston, M. J.; Annan, R. S. Anal. Biochem. 1996, 239, 180192. (9) Medzihradszky, K. F.; Besman, M. J.; Burlingame, A. L. Anal. Chem. 1997, 69, 3986-3994. (10) Ku ¨ ster B.; Wheeler, S. F.; Hunter, A. P.; Dwek, R. A.; Harvey, D. J. Anal. Biochem. 1997, 250, 82-101. (11) Cao, P.; Stults, J. T. Rapid Commun. Mass Spectrom. 2000, 14, 1600-1606. (12) Stensballe, A.; Andersen, S.; Jensen, O. N. Proteomics 2001, 1, 207-222. (13) Ficarro, S. B.; McCleland, M. L.; Stukenberg, P. T.; Burke, D. J.; Ross, M. M.; Shabanowitz, J.; Hunt, D. H.; White, F. M. Nat. Biotechnol. 2002, 20, 301-305. (14) Vacratsis, P. O.; Phinney, B. S.; Gage, D. A.; Gallo, K. A. Biochemistry 2002, 41, 5613-5624. (15) Huddleston, M. J.; Bean, M. F.; Carr, S. A. Anal. Chem. 1993, 65, 877884. (16) Carr, S. A.; Huddleston, M. J.; Bean, M. F. Protein Sci. 1993, 2, 183-186. 10.1021/ac035427d CCC: $27.50
© 2004 American Chemical Society Published on Web 05/01/2004
MS to selectively detect glycopeptides8,15,16 and phosphopeptides21-23 in complex mixtures by inducing formation of low-m/z marker ions indicative of these modifications. For glycosylation, these marker ions include m/z 204 (HexNAc+), 274 (NeuAc - H2O+), 292 (NeuAc+), and 366 (Hex - HexNAc+). Marker ions specific to the modification are generated by collision-induced dissociation (CID) in the source region of a mass spectrometer during chromatographic separation of a complex mixture or single protein digest. These ions can also be produced in the collision cell of QqTof instruments during LC-ESMS.24 Real-time monitoring of the marker ions is used to trigger fraction collection to obtain fractions highly enriched in the glycopeptides. These fractions are further characterized by MS to define each specific attachment site and to determine the heterogeneity of glycoforms present at each site. The instruments most often used for the selective detection of marker ions have been the single and triple quadrupole mass spectrometers. However, any mass spectrometer capable of generating these low-mass ions in the source region can be used for these studies. 3D ion traps have had a major impact in proteomic studies, including “shotgun sequencing” and analysis of posttranslational modifications.25-28 In this report, we describe the optimization and application of selective detection of N-linked glycans on a 3D ion trap mass spectrometer using human R-1 acid glycoprotein and a therapeutic monoclonal antibody for glycopeptide mapping. EXPERIMENTAL SECTION Chemicals. Human R-1 acid glycoprotein (AGP) was purchased from CalbioChem (LaJolla, CA). Neuramidase (cloned from Clostridium perfringens and overexpressed in Escherichia coli) was purchased from New England Biolabs (Beverly, MA). Peptide N4-acetyl-β-glucosaminyl (PNGase F) from Flavobacterium meningosepticum was purchased from New England Biolabs (Beverly, MA). Proteolytic enzymes were purchased from Promega (modified trypsin; Madison, WI) and from Roche Diagnostics (chymotrypsin and Glu-C; Basil, Switzerland). The recombinant therapeutic antibody used was produced in Chinese hamster ovary (CHO) cells and purified as a reagent for our biotherapeutic program. Enzymatic Digestions. All protein digestions were performed in 50 mM NH4HCO3, pH 7.5, using an enzyme/substrate ratio (17) Greis, K. D.; Hayes, B. K.; Comer, F. I.; Kirk, M.; Barnes, S.; Lowary, T. L.; Hart, G. W. Anal. Biochem. 1996, 234, 38-49. (18) Mazsaroff, I.; Yu W.; Kelley, B. D.; Vath, J. E. Anal. Chem. 1997, 69, 25172524. (19) Neubauer, G.; Mann, M. Anal. Chem. 1999, 71, 235-242. (20) Steen, H.; Ku ¨ ster, B.; Fernandez, M.; Pandey, A.; Mann, M. Anal. Chem. 2001, 73, 1440-1448. (21) Huddleston, M.; Annan, R. S.; Bean, M. F.; Carr, S. A. J. Am. Soc. Mass Spectrom. 1993, 4, 710-717. (22) Annan, R, S.; Huddleston, M. J.; Verma, R.; Deshaies, R. J.; Carr S. A. Anal. Chem. 2001, 73, 393-404. (23) Zappacosta, F.; Huddleston, M. J.; Karcher, R. L.; Glefand, V. I.; Carr, S. A.; Annan, R. S. Anal. Chem. 2002, 74, 3221-3231. (24) Bateman, R. H.; Carruthers, J. B.; Hoyes, C.; Jones, J. I.; Langridge, A.; Millar, J. P.; Vissers, C. J. Am. Soc. Mass Spectrom.. 2002, 13, 792-803. (25) Oda, Y.; Huang, K.; Cross, F. R.; Cowburn, D.; Chait, B. T. Proc. Natl. Acad. Sci., U.S.A. 1999, 96, 6591-6596. (26) Zhou, H.; Watts, J. D.; Aebersold, R. Nat. Biotechnol. 2001, 19, 375-378. (27) Washburn, M. P.; Wolters, D.; Yates, J. R., III. Nat. Biotechnol. 2001, 19, 242-247. (28) Zhou, Z.; Licklider, L. J.; Gygi, S. P.; Reed, R. Nature 2002, 419, 182-185.
based on manufacturer’s recommendations for each enzyme used. Reduction was performed using 10 mM dithiothreitol for 1 h at 60°C followed by alkylation with 30 mM iodoacetamide for 1 h in the dark at room temperature. Enzymatic digestions were done at 37 °C for 16 h, and reactions were stopped by acidification with acetic acid. Proteins were proteolytically digested in separate experiments with multiple enzymes to ensure 100% recovery of all five N-linked glycopeptides. For some AGP analyses, Nacetylneuraminic acid (sialic acid) present on AGP was removed by treatment with neuraminidase. For these experiments, AGP was solubilized in 50 mM sodium citrate, pH 6.0, and the digestion was allowed to proceed for 16 h at 37 °C. Postdigestion and prior to LC/MS, protein and enzyme were dialyzed against 50 mM ammonium bicarbonate, pH 8.0, for 16 h to remove any low molecular weight impurities that could interfere with electrospray analysis. In addition, sample was heated to 80 °C for 10 min to inactivate any residual neuraminidase activity that would interfere with chromatography during LC-ESMS. To determine the site of attachment of carbohydrate in each glycopeptide, fractions containing glycopeptides (based on marker ion detection) collected during LC/MS experiments were digested with PNGase F to release N-linked carbohydrates from the peptides and simultaneously convert the attachment site Asn to Asp. Fractions were adjusted to pH 8.6 with a concentrated stock of ammonium bicarbonate. Digestion with PNGase F was performed for 3-16 h at 37 °C utilizing 0.5 units of enzyme/nmol of protein. HPLC. All solvents were HPLC grade (J T Baker; Phillipsburg, NJ). Enzymatically digested proteins were fractionated on a 300 µm i.d. × 15 cm L, PepMap C18 column (Dionex, Marlton, NJ). Chromatographic separations were done with an Agilent 1100 binary pump system (Palo Alto, CA). Fractionation was done using a linear gradient of 2-60% acetonitrile in 0.1% formic acid in 30 min followed by 60-80% acetonitrile in 0.1% formic acid in 5 min. Using microtight fittings (Upchurch, Oak Harbor, WA) and fused silica (Polymicro Technologies, Phoenix, AZ), a precolumn split ratio of 1:150 yielded a column flow rate of 4 µL/min and a postcolumn split ratio of 1:16 yielded a flow rate of 250 nL/min to the mass spectrometer (LCQ Deca ion trap, ThermoFinnigan, San Jose, CA). Fused silica with an inner diameter of 50 µm was used for flow to the mass spectrometer and for waste lines. Fused silica with an inner diameter of 75 µm was used as the fraction collection line. The divert valve supplied with the LCQ Deca was utilized for injections. At the start of the method, the divert valve was set to the inject position. At 10 min, the divert valve was switched to the load position, minimizing delay volumes. One-minute fractions were collected manually in 0.5-mL polypropylene tubes containing 3 µL of 50% acetonitrile during the LC-ESMS run. Fractions shown to contain glycopeptides were further analyzed to identify the site of carbohydrate attachment and to determine the molecular weights of the intact glycopeptides. Mass Spectrometry. All LC-ESMS and LC-ESMS/MS experiments were performed in positive ion mode on a ThermoFinnigan LCQ Deca ion trap mass spectrometer equipped with a custom-designed low-flow electrospray ionization interface (James A. Hill Instrument Services, Arlington, MA). The spray tip employed was an 8-µm-i.d. PicoTip from New Objective (Woburn, MA). Source parameters that had the most impact on improving Analytical Chemistry, Vol. 76, No. 11, June 1, 2004
3113
the signal-to-noise ratio of the carbohydrate marker ions (m/z 204, 274, 292, and 366 ( 0.5) were the tube lens offset, capillary voltage, temperature of the heated capillary, and application of in-source CID. Employing in-source CID increases the ion velocity between capillary and skimmer region in the ion source to cause dissociation. Standard operating conditions for LC-ESMS and LCESMS/MS of peptides typically used a tube lens offset of 5-15 V, capillary voltage of 20-45 V, capillary temperature of 130 °C, and in-source CID set to off. These conditions can vary depending upon the cleanliness of the instrument and generally are optimized by infusion of a standard peptide. Optimization for carbohydrate marker ions (m/z 204 and 366) was performed by infusion of a carbohydrate standard (Man9GlcNAc2; CalbioChem). Source settings found to be optimal for generation of carbohydrate marker ions were as follows: tube lens offset of 100 V; capillary voltage of 75 V; capillary temperature of 175 °C; in-source CID set to on and CE set to 35%. LC-ESMS experiments employing selective detection of carbohydrate marker ions contained two alternating scan events in the experiment method on the LCQ Deca. The first scan event was a full MS scan with a mass range of 400-2000 and in-source CID set to off. The second scan event was a full MS scan with a mass range of 200-400 and in-source CID set to on (CE ) 35%). Tube lens offset, capillary voltage, and capillary temperature were set at the elevated settings described above for both scan events. Maximum injection time and microscan count were set to 300 ms and 3, respectively, for a cycle time of 1.8 s. For the highest sensitivity, m/z 204, 292, and 366 were monitored using single ion monitoring (SIM). To determine site of attachment of the N-linked glycopeptides, one-third of each glycopeptide-containing fraction collected was treated with PNGase F as described above and analyzed by matrixassisted laser desorption/ionization time-of-flight mass spectrometry (MALDI-TOF) to obtain the mass of the deglycosylated peptide, and by LC-ESMS/MS to establish sequence. For MALDI analysis, samples were desalted post PNGaseF digestion using C18 Ziptips (Millipore Inc., Billerica, MA) according to supplied protocols. One microliter of the desalted fractions was analyzed on a Voyager DE-STR mass spectrometer (Applied Biosystems, Framingham, MA) equipped with a N2 laser. Samples were mixed 1:1 with matrix (dihydroxybenzoic acid, 10 mg/mL in 80% acetonitrile/1.0% TFA) and were acquired in reflector mode with the laser voltage between 1800 and 2000 V and a mass range of m/z 600-3000. For LC-MS/MS, sample was analyzed on a 75µm capillary column packed in-house with Vydac C18 reversedphase material (The Nest Group; Southborough, MA). The LCQ Deca was operated with standard source settings and datadependent acquisition. To determine the molecular weight of intact glycopeptides, the remaining material collected during LC-ESMS employing selective detection of carbohydrate marker ions was analyzed by nanospray on a Micromass Q-TOF I mass spectrometer (Micromass, Manchester, U.K.), equipped with a standard Z spray ionization source. One microliter of the collected fractions was loaded into a 2-µm-i.d. gold-coated borosilicate glass nanospray needle (New Objective. Woburn, MA). Samples were acquired in MS mode with a capillary voltage of 1200-1800 V, a cone voltage of 30 V, and a mass range of m/z 450-2000. 3114
Analytical Chemistry, Vol. 76, No. 11, June 1, 2004
RESULTS AND DISCUSSION Human AGP is a heavily glycosylated serum protein with oligosaccharide chains accounting for nearly half of its molecular weight.3,29 AGP has five N-linked glycosylation sites, with complex sugars having bi-, tri-, or tetraantennary structures.30,31 Due to the complexity of glycans, AGP is an excellent standard protein with which to develop and validate methodology that selectively detects glycopeptides during an LC-ESMS analysis of enzymatically digested glycoproteins. We performed several combinations of enzymatic digests of AGP in order to obtain optimal size glycopeptides for HPLC separation and characterization and that would also yield a single site of glycosylation per peptide. Endoproteinase Glu C (V8) gave a reasonable cleavage pattern for most of the N-linked glycosylation sites and was chosen for the analyses described here. Chymotrypsin and trypsin digests were also analyzed by the methods described here (data not shown). In addition to the LCMS analysis of glycopeptides with retention of sialic acid (see Experimental Section), 2,6-linked sialic acids were released from AGP by digestion with neuraminadase prior to proteolysis and LC-MS analysis. Removing the sialic acids decreases the heterogeneity of the glycopeptides and improves reversed-phase separation and electrospray ionization of the various glycoforms. Generation of Marker Ions. Optimal production of carbohydrate marker ions in the source region of the LCQ Deca requires more than simply turning on the in-source CID parameter. Specifically, it is necessary to increase tube lens offset, capillary temperature, and capillary voltage (see Experimental Section) during the LC-ESMS analysis of digest mixtures of glycoproteins. To maximize formation of these ions in these experiments, two MS scan events were employed. The first MS scan event included the mass range 200-400 with application of in-source CID (see Experimental Section) to generate the carbohydrate-specific marker ions by fragmenting glycopeptides present A limited mass range is used to improve sensitivity for detection of the marker ions and because the higher mass region would be dominated by fragments of the peptides under CID-on conditions. Maximum sensitivity for detection of the marker ions is obtained using SIM (see below). In the second MS scan event, the mass range 400-2000 was acquired with in-source CID set to off to obtain a TIC trace and to attempt to produce spectra predominantly exhibiting molecular weight-related ions of the peptides and glycopeptides present during that scan cycle. However, in the current version of the Xcalibur software, it is not possible to change source conditions, such as the tube lens offset, capillary voltage, or capillary temperature, while toggling between scan events within the same LC segment. Therefore, these specific source parameters were maintained at elevated values during this second scan event, and considerable parent dissociation is observed for peptides and glycopeptides. Extracted ion chromatograms of m/z 204 (HexNAc+), 274 ([NeuAc - H2O]+); 292 (NeuAc+), and 366 (Hex - HexNAc+) were generated by the data system to determine which regions (29) Treuheit, M. J.; Costello, C. E.; Halsall, H. B. Biochem. J. 1992, 283, 105112. (30) Fournet, B.; Montreuil, J.; Strecker, G.; Dorland, L.; Haverkamp, J.; Vliegenthart, J. F. G.; Binette, J. P.; Schmid, K. Biochemistry 1978, 17, 52065214. (31) Yoshima, H.; Matsumoto, A.; Mizuochi, T.; Kawasaki, T.; Kobata, A. J. Biol. Chem. 1981, 256, 8476-8484.
Figure 1. TICs and XICs of m/z 204, 292, and 366 for three LCESMS analyses of Glu C digestions of AGP. (A) Data obtained using standard ion source conditions for LC-ESMS on the LCQ without in-source CID. (B) In-source CID on and other ion source conditions optimized for the production of carbohydrate marker ions. (C) AGP treated with neuramidase prior to proteolysis and LC-ESMS; same instrument conditions as in (B). The y axis for each XIC of the same ion has been scaled according to the highest observed intensity for that ion among the three experiments. Complexity of XICs in panel B is a result of various carbohydrate isoforms present. Gradient conditions employed differ slightly for all three experiments.
of the chromatogram (and which of the fractions being collected simultaneously) contained the glycopeptides. To have confidence that a glycopeptide is present, abundant ions of m/z 204 and 366 must maximize simultaneously in the extracted ion current traces and corresponding mass spectra. Detection of NeuAc+ (with or without the neutral loss of H2O) simultaneously with the HexNAc+ and Hex - HexNAc+ marker ions indicates the presence of sialic acid. Figure 1 displays the total ion chromatograms (TICs) and extracted ion chromatograms (XICs) from three separate LCESMS analyses of Glu C digests of AGP. The extracted ion chromatograms for each experiment have been normalized to each other in order to compare the appearance or disappearance of the specific marker ions. Panel A data were produced using
standard ion source conditions for LC-ESMS on the LCQ, and in-source CID was not employed to purposefully generate markerions. The extracted ion chromatograms for m/z 204, 292 and 366 (Figure 1A, traces 2-4) are fairly weak and noisy, and they do not exhibit peaks that are strongly correlated, a necessary criterion for indicating the presence of glycopeptides. In contrast, when LC-ESMS conditions are optimized for the detection of carbohydrate marker ions, the extracted ion chromatograms (Figure 1B) clearly show the presence of glycopeptides in a complex protein digest of AGP. Marker ions m/z 204, 292, and 366 are more abundant than the corresponding traces observed in Figure 1A, and these ions simultaneously maximize in a number of discrete locations in the chromatogram. The presence of m/z 292 (Figure 1B-3) coincident with m/z 204 and 366 (Figure 1B-2 and Figure 1B-4, respectively) confirms that sialic acid is present on the glycopeptides present in these peaks. This observation is consistent with literature results29 and with the fact that no neuramidase treatment was performed prior to digestion for the LC-ESMS experiment shown in Figure 1B. Figure 1C shows the selective detection of carbohydrate marker ions for a sample of AGP that was treated with neuramidase prior to digestion with Glu C. In this experiment, m/z 292 is not detected (Figure 1C-3), indicating complete removal of the sialic acids. Removing sialic acid greatly decreased the glycopeptide heterogeneity as evidenced by the much smaller number of discrete glycopeptide peaks observed. This is in part due to the collapse of glycoforms with the same peptide sequence. Confirmation of Attachment Sites and Determination of Glycan Structure. During the LC-ESMS analyses, ∼2/3 of the eluant from the 300-µm-i.d. RP column was collected automatically in 1-min fractions. Fractions identified by marker ions as containing glycopeptide(s) were further analyzed. To determine site of attachment of carbohydrate in the N-linked glycopeptides present, aliquots of the collected fractions were treated with PNGase F. This endoglycosidase cleaves the amide bond formed between the side chain of Asn and the ultimate GlcNAc residue of the carbohydrate chain, releasing the entire carbohydrate from the peptide and simultaneously converting the attachment site asparagine to aspartic acid.32,33 The amino acid sequence of the peptide can then be determined by mass if the protein sequence is known. This is illustrated in Figures 2A and 3A, which show the MALDI mass spectra for two glycopeptide-containing fractions from Figure 1C eluting at 39 and 43 min, respectively. Peaks observed in the MALDI mass spectrum shown in Figure 2A correspond to the predicted peptide masses of the Glu C peptide Ile62-Glu75 of AGP and a missed cleavage product corresponding to Ile62-Asp76 (both masses shifted upward by 1 Da due to the conversion of the attachment-site Asn to Asp upon release of the carbohydrate), confirming that this part of the protein sequence contains an N-linked carbohydrate. Carbohydrate is presumably attached to Asn72 (now Asp72), as this residue is in the known consensus sequence for attachment of N-linked sugar, Asn-X-Ser/Thr (where X is any residue except proline). Similarly, the mass peak observed in Figure 3A corresponds to Glu C peptide Tyr83-Glu102, with the site of attachment at Asn93 (now Asp93). As expected, the (32) Tarantino, A. L.; Gomez, C. M.; Plummer Jr., T. H. Biochemistry 1985, 24, 4665-4671. (33) Carr, S. A.; Barr J. R.; Roberts, G. D.; Anumula, K. R.; Taylor, P. B. Method Enzymol. 1990, 193, 501-27.
Analytical Chemistry, Vol. 76, No. 11, June 1, 2004
3115
Figure 2. (A)MALDI-TOF mass spectrum of glycopeptide-containing fraction eluting at 39 min from Figure 1C following PNGase F digestion to remove carbohydrate. Peptides shown are the deamidated (*) sequences derived from AGP that form upon PNGase F treatment. Data indicate site of carbohydrate attachment to be Asn72. (B) Electrospray mass spectrum of glycopeptide-containing fraction eluting at 39 min from Figure 1C. Suggested composition of the glycoforms are indicated and based upon the intact molecular weight of the glycopeptide and the deglycosylated peptide shown in (A).
observed masses in Figures 2A and 3A are increased by 1 Da compared to the predicted masses of the former glycosylated peptides due to deamidation of Asn. In case of ambiguity, the sequences of the peptides and sites of attachment can be further corroborated by sequencing the peptides by MS/MS. Once the identity (and therefore the mass) of the attachment site peptide is established, it is then possible to extract glycoform composition from the masses of the intact glycopeptides as previously described. 16,31 Figures 2B and 3B show the intact glycopeptide mass analysis of two fractions (eluting at 39 and 43 min, respectively, from Figure 1C) from a Glu C digest of AGP following removal of sialic acid with neuramidase. Aliquots of the collected glycopeptide-containing fractions were analyzed by nanoelectrospray on a Micromass Q-TOF mass spectrometer. Glycopeptides in these fractions were observed as triply charged ions in the electrospray mass spectra. Only the glycosylated peptides are labeled, although the spectra contain peaks arising from peptides and alkali-metal cationized forms of the glycopeptides as well. The differences in mass between the peptide (attachment site, Asn) and the various glycoforms of the glycopeptides correspond to the in-chain masses of the respective carbohydrates attached in terms of hexose, N-acetylhexosamine, deoxyhexose, and N-acetylneuramic acid. The attachment-site 3116 Analytical Chemistry, Vol. 76, No. 11, June 1, 2004
Figure 3. (A) MALDI-TOF mass spectrum of glycopeptide-containing fraction eluting at 43 min from Figure 1C following PNGase F digestion to remove carbohydrate. Peptide shown is the deamidated (*) sequence derived from AGP that formed upon PNGase F treatment. Data indicate site of carbohydrate attachment to be Asn93. (B) Electrospray mass spectrum of glycopeptide-containing fraction eluting at 43 min from Figure 1C. Suggested compositions of the glycoforms are indicated and based upon the intact molecular weight of the glycopeptide and the deglycosylated peptide shown in (A).
sequences and compositions of carbohydrates determined for each of the five N-linked sites in AGP are summarized in Table 1. The structures shown in the table and on the spectra are putative and are based on the known biosynthetic pathways for carbohydrates and the common structures present on glycoproteins.2 N-Linked carbohydrates synthesized in mammalian cells generally contain a common core structure consisting of two GlcNAc attached to a trimannose and further extended by branching antennas, consisting of Gal, GlcNAc, sialic acid, and sometimes fucose.2 Either ESI-MS or MALDI-TOF can be used to analyze glycopeptides for intact molecular weight. However, the presence of non-glycopeptides, of sialic acid, and the heterogeneity of the glycoforms limit the sensitivity for analysis of glycopeptides by MALDI. Resolution is also affected giving broad unresolved peaks and, therefore, inaccurate MW determination. ESMS appears to be much better at resolving the complex glycoforms, is less prone to suppression, has higher intrascan dynamic range, and has the ability to characterize fractions containing two or more sites. Analysis of Therapeutic Monoclonal Antibody. Many biotherapeutics currently under development in the pharmaceutical industry are monoclonal antibodies, and characterization of the site and structures of carbohydrates present is required by
Table 1. Summary of Reported and Observed Glycan Structures for AGP site
sequence enzyme
(Asn33)
LVPVPITNATLDR trypsin/V8
(Asn56)
RNEEYNKSVQEIQATF chymotrypsin
(Asn72)
IQATFFYFTPNTE Glu C (V8)
(Asn93)
YQTRQDQCIYNTTYLNVQRE Glu C (V8)
(Asn103)
NTISRYVGGQE Glu C (V8)
reported glycan structurea
calculated MH+b
observed MH+c
biantennary triantennary triantennary + fucose biantennary triantennary tetraantennary biantennary triantennary tetraantennary + fucose tetraantennary tetraantennary + fucose triantennary triantennary + fucose tetraantennary tetraantennary + fucose tetraantennary + 2 fucose tetraantennary+GalGlcNAc biantennary triantennary tetraantennary + fucose tetraantennary tetraantennary + fucose tetraantennary+2 fucose
3031.3 3396.5 3542.5 3578.5 3943.6 4308.8 3329.4 3694.5 3840.6 4059.6 4205.7 4580.9 4726.9 4946.0 5092.1 5238.1 5311.1 2903.2 3268.3 3414.3 3633.4 3779.5 3925.6
ndd 3396.5 3542.6 3578.4 3943.7 4310.7 3329.7 3694.6 3840.5 4059.8 4206.0 4581.4 4727.1 4946.6 5093.4 5239.6 5312.9 nd 3268.5 3414.6 3633.6 3779.7 3924.6
a Reported glycan structure from ref 29. b Calculated MW based on reported glycan structures. c Deconvoluted MW based on observed m/z from Q-TOF.
Figure 4. (A) TIC for tryptic digest of therapeutic monoclonal antibody acquired under conditions optimized for selective detection of carbohydrate marker ions. XICs of m/z 204 (B), 292 (C), and 366 (D) indicating elution of glycopeptide at 26 min. Absence of signal for m/z 292 in (C) at 26 min indicates lack of sialic acid on the glycopeptide.
regulatory agencies for drug approval. Therefore, as useful realworld application, we applied the methodology described above to locate and characterize glycopeptides in digests of a therapeutic monoclonal antibody produced in CHO cells. Based upon the consensus sequence for N-glycosylation, there is one N-linked glycosylation site at Asn296 of the antibody sequence. This residue is predicted to be found in the tryptic peptide Glu293-Arg301. Therefore, trypsin was chosen as the initial enzyme to cleave with, and the reduced, alkylated and trypsinized antibody was analyzed by LC-ESMS with carbohydrate marker ion detection and online fraction collection as detailed for AGP, above. Figure 4A shows the TIC and extracted ion chromatograms for m/z 204 (Figure 4B), 292 (Figure 4C), and 366 (Figure 4D). Both m/z 204 and 366 comaximize at the peak eluting at 26.6 min, indicating that
Figure 5. Q-TOF electrospray mass spectrum of glycopeptidecontaining fraction at 26 min from Figure 4B. Ions observed in this spectrum are doubly charged and indicate the presence of different glycoforms, which are calculated based upon the intact molecular weight of the glycopeptide and the confirmed mass (data not shown) of the deglycosylated peptide.
the tryptic glycopeptide is contained in this fraction. The absence of a peak in the extracted ion chromatogram for m/z 292 (Figure 4C) at 26.6 min indicates that the carbohydrate of the glycopeptide has little or no sialic acid. The composition of the carbohydrate on the glycopeptide eluting at 26.6 min was defined as described above for AGP. Figure 5 shows the electrospray mass spectrum acquired on a Q-TOF mass spectrometer and the predicted carbohydrate structures based on the determined masses of the carbohydrate chains. There is no evidence of sialylation, which is consistent with the marker ion data. The structures shown are consistent with published results for a recombinant antibody to respiratory syncytial virus.34 The putative glycoforms have a complex, biantennary structure with core fucosylation. Analytical Chemistry, Vol. 76, No. 11, June 1, 2004
3117
Sensitivity. The sensitivity for detection and collection of glycopeptides using the LCQ Deca was evaluated using SIM of a tryptic digest of fetuin, another well-characterized glycoprotein. The sensitivity achievable for marker ion detection is limited by the HPLC column diameter and the need to employ flow rates consistent with the ability to collect fractions for further analysis as has been discussed previously. 23 Using a 300-µm-i.d. reversedphase column and a flow rate of 4 µL/min, as little as 25 pmol of a fetuin digest could be injected to selectively detect glycopeptides with a S/N of at least 5:1. We have also been able to employ 180µm i.d. at flow rates of 2 µL/min to detect marker ions by SIM and collect the corresponding fractions injecting as little as 10 pmol of a glycoprotein digest (data not shown). Smaller diameter columns and lower flow rates more optimal for the column diameter being used (e.g., 200 nL/min for a 75-µm-i.d. column) can be used for detection of glycopeptides in complex digests at subpicomole levels, but the nanoliter flow rates used with these columns make on-line fraction collection extremely difficult. CONCLUSIONS We have demonstrated in this report the capability of routinely using 3D ion trap mass spectrometers for the selective detection of carbohydrate marker ions with simultaneous, on-line fraction collection. The method is simple to implement on LCQ ion traps and greatly facilitates identification of glycopeptides for further analysis. Optimal in-source CID conditions were established for (34) Roberts, G. D.; Johnson, W. P.; Burman, S.; Anumula, K. R.; Carr, S. A. Anal. Chem. 1995, 67, 3613-3625.
3118
Analytical Chemistry, Vol. 76, No. 11, June 1, 2004
analysis of enzymatically digested glycoproteins during LCESMS. In addition to enabling in-source CID, it was essential to also increase the tube lens offset, capillary voltage, and capillary temperature to obtain optimal signal-to-noise ratios for marker ions m/z 204, 292, and 366. We have also determined that carbohydrate marker ions can also be generated on newer model LCQs such as the ThermoFinnigan Deca XP and Deca XPplus (data not shown). However, we found that that optimal source conditions are different for those instruments, most likely due to the changes in ion optics as compared to the Deca. In principle, it should also be possible to employ the methods outlined here to optimize and use the glycopeptide selective marker ion strategy on other vendor’s 3D ion traps as well as on 2D ion trap instruments. Current sensitivity for detection and collection of glycopeptides using 180-µm-i.d. columns requires a minimum ∼10 pmol of glycoprotein digest. Subpicomole sensitivity for selective detection of glycopeptides should be possible using smaller i.d. columns at lower flow rates (e.g., 75-µm-i.d. columns at 200 nL/min). Such experiments would facilitate comparison of the glycosylation pattern (stoichiometry, different sites) of a glycoprotein expressed under differing cell growth and culture conditions or for differential analysis of the “glycome” of complex mixtures of glycoproteins derived from a cell or tissue under differing stimulation or stress conditions.
Received for review December 3, 2003. Accepted March 15, 2004. AC035427D
