- Email: [email protected]
Detailed Characterization of Carbohydrate Linkage and Sequence in an Ion Trap Mass Spectrometer: Glycosphingolipids
ANALYTICAL BIOCHEMISTRY ARTICLE NO.
259, 28 –33 (1998)
AB982619
Detailed Characterization of Carbohydrate Linkage and Sequence in an Ion Trap Mass Spectrometer: Glycosphingolipids Vernon N. Reinhold1 and Douglas M. Sheeley* Department of Microbiology, Boston University School of Medicine, 715 Albany Street, Boston, Massachusetts 02118; and *Glaxo Wellcome Inc., Research Triangle Park, North Carolina
Received December 1, 1997 DEDICATED TO ROGER JEANLOZ IN HONOR OF HIS 70TH BIRTHDAY
Electrospray ionization with a quadrupole ion trap (qIT) mass analyzer has been utilized to ascertain structural detail obtained with glycoconjugate samples. In this report, an isomeric disialyl glycosphingolipid sample extracted from human brain tissue was evaluated for sequence, branching, and linkage information. Results were obtained that were qualitatively comparable with triple-quadrupole instruments (Q1q2Q3) with major carbohydrate fragments from C1-O glycosidic rupture and additional fragments that provided a determination of sphingosine and N-acyl heterogeneity of the ceramide moiety. In unique contrast, however, the qIT extended carbohydrate understanding through multistep mass spectrometric (MSn) studies providing for the first time pyran cross-ring cleavages that define the interresidue linkage structure for glycolipids. This was achieved by selecting secondary fragments (MS2) free from the energy sinks of facile bond rupture that dominate product ion spectra. Isolation and activation of these substructures result in a more uniform distribution of fragments, providing structural insights previously inaccessible by tandem mass spectrometry. © 1998 Academic Press
1 To whom correspondence should be addressed at Department of Microbiology, Mass Spectrometry Resource, Boston University School of Medicine, 715 Albany St., Boston, MA 02118. Fax: (617) 638-6761. E-mail: [email protected]. 2 Abbreviations used: ES, electrospray; ITMS, quadrupole ion trap mass spectrometry; CID, collision-induced dissociation; MSn, notation for multistep isolation and resonance ejection in an ion trap mass spectrometer; GSL, glycosphingolipid; DMSO, dimethyl sulfoxide.
28
Developments in generating ions from a condensed phase have brought new dimensions to our understanding of biopolymer structure. Problems, inconceivable a few short years ago, turn out now to be routine and much of that credit goes to the techniques of electrospray (1) ionization and matrix-assisted laser desorption (2). As applied to biopolymers, the first established success of ES2 was realized with polar molecules ionized from aqueous aerosols, a development highly suited to peptide sequencing when followed by collisional activation. In contrast to this past and important focus, we have emphasized the value of ES when applied to lipophilic samples evaporated from less polar solvents, a strategy concordant with the detailed analysis of glycoconjugates. In this approach the neutral samples were ionized by sodium adduction in ES which provides a profile of intact components. These molecular-weight-related ions were selected in Q1 for low-energy collision in q2 and the products mass analyzed in Q3 (where Q1 and Q3 are mass filters and q2 is a RF-only collision cell). For glycosphingolipid (GSL) samples structural information was obtained in single-bond glycosidic cleavage ions. Sequence information was found in both ‘‘reducing’’ and ‘‘nonreducing’’ fragments, while branching information was retained in reducing-end or ceramide-containing ions (3–5). These structural elements were also observed within internal fragments, as a consequence of a secondary terminal loss. This combination of mass profiling and collisionally induced fragmentation has allowed characterization of most GSL samples, but has relied heavily on anticipated structural motifs, including linkage and moiety position. This reliance on motifs for linkage has been necessary because fragments that define linkage are not 0003-2697/98 $25.00 Copyright © 1998 by Academic Press All rights of reproduction in any form reserved.
MULTISTEP SPECTROMETRIC ANALYSIS OF GLYCOSPHINGOLIPIDS
29
FIG. 1. Electrospray ionization ion trap mass spectrometry component profile of a human brain disialoganglioside sample isolated from human brain tissue. Purified by thin-layer chromatography.
detected in the presence of certain extremely labile glycosidic bonds (e.g., neuraminyl ketosides and 2-acetamidohexosamine glycosides), which rupture to generate abundant fragments and dominate the spectrum. In this study we utilize a multistep excitation strategy, MSn, provided by quadrupole ion trap mass spectrometry (ITMS), to isolate and determine structural details from these labile fragments that dominate single dimensional experiments. Imparting resonance excitation to these smaller pieces has provided new, previously unobserved, fragmentation pathways for an improved structural understanding of GSL samples. EXPERIMENTAL
Instrumentation. The ITMS used in this study was a Finnigan-MAT LCQ (Finngan Corp., San Jose, CA), coupled with an ESI source (6) through a vacuum chamber that allows for differential pumping between the ion source and the analyzer region. Methylated samples were directly infused at 1.5 ml/min from a solution of 1 mM sodium acetate in 70/30 MeOH/H2O. Ions are injected axially into the ion trap by a gate lens and a trapping field was established with a 100- to 1100-kHz radio frequency applied to the ring electrode. Several excellent reviews of ITMS operation have recently been reported (7–9).
Methylation. Vacuum-desiccated samples isolated and extracted from HRTLC plates were dissolved in 200 ml of a NaOH/DMSO suspension, prepared by vortexing DMSO and powdered sodium hydroxide. After 1 h at RT, 50 ml of methyl iodide was added and the solution was incubated for 1 h at RT with occasional vortexing (10). Samples were partitioned by adding 1 ml of chloroform and the suspension back extracted four times with 2–3 ml of 30% acetic acid and the chloroform layer was taken to dryness and stored at -20°C. For more complex extracts methylation was repeated and the chloroform layer was backwashed with water. Thin-layer chromatographic fractions were scraped from the plate, directly methylated in the presence of silica, and then extracted. RESULTS AND DISCUSSION
Figure 1 shows a molecular weight profile obtained from an HRTLC spot following extraction and methylation. The two major components, m/z (1118.1)21 and (1131.7)21, support the expectation of disialyl GSLs and their mass difference indicates the ions to be doubly charged. The doublet is characteristic of sphingosine (C18) and eicosasphingosine (C20) heterogeneity located in the ceramide moiety and the profile suggests the structures to be isomeric with either GD1a
30
REINHOLD AND SHEELEY
FIG. 2. Disialoganglioside fragmentation by MS2; isolation and resonance ejection of the doubly charged ion, m/z 1131.721 from Fig. 1.
or GD1b. Several minor components were also resolved in this extract and their molecular-weight-related profiles indicate them to be GD2, GM3, GA1, GM2, and GD3. The amounts of these materials were insufficient for confirmation. Ion selection of the major component, (1131.7)21, was accomplished by scanning the amplitude of the fundamental RF voltage in a reverse-then-forward manner while applying a resonance signal which ejected all ions greater or less than m/z 1131.7 (61 Da). Fragmentation of the selected ion was induced by resonance excitation (a ‘‘tickle’’ voltage) which alters the ion trajectory from the center of the trap causing CID with the helium damping gas. The product ion spectrum is presented in Fig. 2. Most of the major ions were comparable to data obtained with a triple-quadrupole instrument, but significant qualitative and quantitative differences indicate the modes of excitation are clearly different. The ceramide- (m/z 604.5) and carbohydrate- (m/z 828.9) containing fragments were easily discernible, although the former was in lower abundance. Fragments for Neu5Ac (m/z 375.9/397.9), a nonreducing terminal Hex-HexN (m/z 486.1), and a tandem-linked Neu5Ac–Neu5Ac (m/z 737.3/759.4) were at comparable intensities. The remaining fragments can be identified as products of one or more lost termini and this spectrum clearly defines a GD1b sequence. An ion (m/z 847.6) strongly indicates that GD1b is not the
only isomer present. This fragment, (Neu5Ac-HexHexNAc z Na)1, argues for a small amount of GD1a, which was also present when the glycan itself was isolated and activated (see below, Fig. 4). However, no linkage information was detected. Selection of the ceramide fragment m/z 604.5 by again ejecting all ions greater and less than m/z 604.5 (by 61 Da) followed by CID with a tickle frequency provided the spectrum in Fig. 3. Methanol elimination (m/z 572.1) supports an original 3-hydroxy eicosasphingenine which again was suggested by the abundant ion at m/z 306.3, a ceramide product following N-acyl elimination. This major fragment strongly suggests the neutral loss N-acyl fatty acid moiety to be stearic acid by mass difference (see inset, Fig. 3). Isolation and activation of the carbohydrate-containing fragment, m/z 828.8, provided linkage information, in addition to the expected sequence and branching detail (Figs. 4 and 5). The selected doubly charged ion expelled a sodium ion upon activation (m/z 1634.9) and also showed losses of single and tandem-linked neuraminyl residues (m/z 641.121/1259.7 and 898.3, respectively). Losses of the two nonreducing terminal fragments, Hex-HexN (m/z 486.0) and the mono- and tandem neuraminyl residues (m/z 375.8/397.7 and 737.3/759.2), supported the proposed earlier GD1b structure (Fig. 2) with fragment m/z 847.6 suggesting a small amount of GD1a. An additional loss of the reduc-
MULTISTEP SPECTROMETRIC ANALYSIS OF GLYCOSPHINGOLIPIDS
31
FIG. 3. Ceramide characterization by MS3; isolation and resonance ejection of the fragment ion, m/z 604.5 from Fig. 3.
FIG. 4. Component carbohydrate characterization by MS4; isolation and resonance ejection of the fragment ion, m/z 828.821 from Fig. 3.
32
REINHOLD AND SHEELEY
FIG. 5.
Scale expansion of Fig. 4 (MS4) for detection of cross-ring fragments.
ing terminal glucosyl, residue with one (m/z 1055.4) or two (m/z 675.9) neuraminyl losses, and fragment m/z 898.3 supports the core sequence of this GSL, HexHexN-(HO)Hex-Hex-OH. Closer examination also indicates an ion 14 Da lower than these fragments (Fig. 5), again supporting a small amount of the GD1a isomer with the same sequence, but with modified Neu5Ac branching [HOHex-HexNAc-(HO)Hex-Hex (e.g., m/z 1040.1 and 884.3)]. Additional structural detail was apparent upon expansion of the low-mass end of the spectrum (Fig. 5) where a series of cross-ring cleavages (or their absence) were detected, suggesting the backbone linkage structure, Hex-3HexN-4(HO)Hex-4HexOH, m/z 574.0 and 1143.5, and the neuminyl linkage, m/z 545.3 (although this fragment would be isobaric for 237, 238, or 239 ketosidic linkages). Additional reducing-end fragments were observed and also losses of nonreducing termini which corroborate the overall glycan structure. Assuming the usual monomer motif for GSL and their anomeric configurations, the structure presented in Scheme 1 can be proposed for the major component. Mass spectral techniques using multiple low-energy collisions provided by Q1q2Q3 instrumentation have proven to be effective strategies for understanding carbohydrate structural detail, e.g., molecular topology (sequence and branching) from single and multiple glycosidic bond cleavage and linkage information from
cross-ring cleavage fragments pendent to nonreducing termini (3, 11, 12). They are particularly abundant in neutral oligomers where collision energy cannot be dissipated easily and are summarized in Scheme 2. Thus, from the nonreducing terminus, cross-linkage fragments were observed for the branched Neu5Ac(28)Neu5Ac linkage and are not expected for the back-
SCHEME 1. Summary of proposed ion fragments determined for the disialoganglioside GD1b from MSn spectra.
MULTISTEP SPECTROMETRIC ANALYSIS OF GLYCOSPHINGOLIPIDS
33
bohydrate sequencing by ITMS need to be clarified and further studied, this preliminary report has identified GSL linkage detail that was not observed with Q1q2Q3 instrumentation. ACKNOWLEDGMENTS
SCHEME 2. Cross-ring cleavage fragments observed here and in neutral oligosaccharides (11, 12) and detected in MS3 and MS4 spectra (Figs. 4 and 5). The masses identified are increments above the normal glycosidic fragment shown in parentheses in Scheme 1.
bone Hex(1-3)HexNAc linkage. Cross-ring fragments were also observed as consistent, albiet minor, fragments for the subsequent HexNAc(1-4)Hex and Hex(1-4)Hex, at m/z 574.0 and 1143.5 (Figs. 5 and 4, respectively). Unfortunately, these structural features are not uniformly detectable and collision spectra of many glycoconjugates are incompletely characterized. A primary constraint is that higher molecular weight samples show a general decrease in the extent of fragmentation with increasing size. Adding further complication is the fact that specific residues (HexNAc, NeuNAc) exhibit facile glycosidic rupture yielding product ions that dominate the spectra at the expense of less abundant, more informative linkage and branching fragments. This study had the expectation that an additional one or two dimensional collision experiments (MS3– 4) could isolate selected components of structure both smaller in size and possibly free from the dominant energy sinks of facile bond cleavage to reveal linkage and branching topology. Although many details of car-
We thank Bruce B. Reinhold for his careful review of the manuscript and offering important suggestions and the Department of Analytical Sciences, Glaxo Wellcome Inc., for use of the LCQ. The human brain disialoglycosphingoside sample was provided by H. C. Yohe, VA Medical and Regional Office Center (White River Junction, VT). This work has been supported in part by NIH Grant R01 GM54045 (V.N.R.).
REFERENCES 1. Fenn, J. B., Mann, M., Meng, C. K., Wong, S. F., and Whitehouse, C. M. (1989) Science 246, 64 –71. 2. Karas, M., and Hillenkamp, F. (1988) Anal. Chem. 60, 2299 – 2301. 3. Reinhold, B. B., Chan, S.-Y., Chan, S., and Reinhold, V. N. (1994) Org. Mass Spectrom. 29, 736 –746. 4. Stroud, M. R., Handa, K., Ito, K., Salyan, M. K., Fang, H., Levery, S. B., Hakomori, S., Reinhold, B. B., and Reinhold, V. N. (1995) Biochem. Biophys. Res. Commun. 209, 777–787. 5. Stroud, M. R., Salyan, M. E. K., Ito, K., Handa, K., Levery, S. B., Hakomori, S.-I., Reinhold, B. B., and Reinhold, V. N. (1996) Biochemistry 35, 758 –769. 6. Whitehouse, C. M., Dreyer, R. N., Yamashita, M., and Fenn, J. (1985) Anal. Chem. 57, 675– 679. 7. Schwartz, J. C., and Jardine, I. (1996) Methods Enzymol. 270, 522–586. 8. March, R. E. (1997) J. Mass Spectrom. 32, 351–369. 9. Jonscher, K. R., and Yates, J. R. (1997) Anal. Biochem. 244, 1–15. 10. Ciucanu, I., and Kerek, F. (1984) Carbohydr. Res. 131, 209 –217. 11. Reinhold, V. N., Reinhold, B. B., and Costello, C. E. (1995) Anal. Chem. 67, 1772–1784. 12. Reinhold, V. N., Reinhold, B. B., and Chan, S. (1996) Methods Enzymol. 270, 377– 402.
259, 28 –33 (1998)
AB982619
Detailed Characterization of Carbohydrate Linkage and Sequence in an Ion Trap Mass Spectrometer: Glycosphingolipids Vernon N. Reinhold1 and Douglas M. Sheeley* Department of Microbiology, Boston University School of Medicine, 715 Albany Street, Boston, Massachusetts 02118; and *Glaxo Wellcome Inc., Research Triangle Park, North Carolina
Received December 1, 1997 DEDICATED TO ROGER JEANLOZ IN HONOR OF HIS 70TH BIRTHDAY
Electrospray ionization with a quadrupole ion trap (qIT) mass analyzer has been utilized to ascertain structural detail obtained with glycoconjugate samples. In this report, an isomeric disialyl glycosphingolipid sample extracted from human brain tissue was evaluated for sequence, branching, and linkage information. Results were obtained that were qualitatively comparable with triple-quadrupole instruments (Q1q2Q3) with major carbohydrate fragments from C1-O glycosidic rupture and additional fragments that provided a determination of sphingosine and N-acyl heterogeneity of the ceramide moiety. In unique contrast, however, the qIT extended carbohydrate understanding through multistep mass spectrometric (MSn) studies providing for the first time pyran cross-ring cleavages that define the interresidue linkage structure for glycolipids. This was achieved by selecting secondary fragments (MS2) free from the energy sinks of facile bond rupture that dominate product ion spectra. Isolation and activation of these substructures result in a more uniform distribution of fragments, providing structural insights previously inaccessible by tandem mass spectrometry. © 1998 Academic Press
1 To whom correspondence should be addressed at Department of Microbiology, Mass Spectrometry Resource, Boston University School of Medicine, 715 Albany St., Boston, MA 02118. Fax: (617) 638-6761. E-mail: [email protected]. 2 Abbreviations used: ES, electrospray; ITMS, quadrupole ion trap mass spectrometry; CID, collision-induced dissociation; MSn, notation for multistep isolation and resonance ejection in an ion trap mass spectrometer; GSL, glycosphingolipid; DMSO, dimethyl sulfoxide.
28
Developments in generating ions from a condensed phase have brought new dimensions to our understanding of biopolymer structure. Problems, inconceivable a few short years ago, turn out now to be routine and much of that credit goes to the techniques of electrospray (1) ionization and matrix-assisted laser desorption (2). As applied to biopolymers, the first established success of ES2 was realized with polar molecules ionized from aqueous aerosols, a development highly suited to peptide sequencing when followed by collisional activation. In contrast to this past and important focus, we have emphasized the value of ES when applied to lipophilic samples evaporated from less polar solvents, a strategy concordant with the detailed analysis of glycoconjugates. In this approach the neutral samples were ionized by sodium adduction in ES which provides a profile of intact components. These molecular-weight-related ions were selected in Q1 for low-energy collision in q2 and the products mass analyzed in Q3 (where Q1 and Q3 are mass filters and q2 is a RF-only collision cell). For glycosphingolipid (GSL) samples structural information was obtained in single-bond glycosidic cleavage ions. Sequence information was found in both ‘‘reducing’’ and ‘‘nonreducing’’ fragments, while branching information was retained in reducing-end or ceramide-containing ions (3–5). These structural elements were also observed within internal fragments, as a consequence of a secondary terminal loss. This combination of mass profiling and collisionally induced fragmentation has allowed characterization of most GSL samples, but has relied heavily on anticipated structural motifs, including linkage and moiety position. This reliance on motifs for linkage has been necessary because fragments that define linkage are not 0003-2697/98 $25.00 Copyright © 1998 by Academic Press All rights of reproduction in any form reserved.
MULTISTEP SPECTROMETRIC ANALYSIS OF GLYCOSPHINGOLIPIDS
29
FIG. 1. Electrospray ionization ion trap mass spectrometry component profile of a human brain disialoganglioside sample isolated from human brain tissue. Purified by thin-layer chromatography.
detected in the presence of certain extremely labile glycosidic bonds (e.g., neuraminyl ketosides and 2-acetamidohexosamine glycosides), which rupture to generate abundant fragments and dominate the spectrum. In this study we utilize a multistep excitation strategy, MSn, provided by quadrupole ion trap mass spectrometry (ITMS), to isolate and determine structural details from these labile fragments that dominate single dimensional experiments. Imparting resonance excitation to these smaller pieces has provided new, previously unobserved, fragmentation pathways for an improved structural understanding of GSL samples. EXPERIMENTAL
Instrumentation. The ITMS used in this study was a Finnigan-MAT LCQ (Finngan Corp., San Jose, CA), coupled with an ESI source (6) through a vacuum chamber that allows for differential pumping between the ion source and the analyzer region. Methylated samples were directly infused at 1.5 ml/min from a solution of 1 mM sodium acetate in 70/30 MeOH/H2O. Ions are injected axially into the ion trap by a gate lens and a trapping field was established with a 100- to 1100-kHz radio frequency applied to the ring electrode. Several excellent reviews of ITMS operation have recently been reported (7–9).
Methylation. Vacuum-desiccated samples isolated and extracted from HRTLC plates were dissolved in 200 ml of a NaOH/DMSO suspension, prepared by vortexing DMSO and powdered sodium hydroxide. After 1 h at RT, 50 ml of methyl iodide was added and the solution was incubated for 1 h at RT with occasional vortexing (10). Samples were partitioned by adding 1 ml of chloroform and the suspension back extracted four times with 2–3 ml of 30% acetic acid and the chloroform layer was taken to dryness and stored at -20°C. For more complex extracts methylation was repeated and the chloroform layer was backwashed with water. Thin-layer chromatographic fractions were scraped from the plate, directly methylated in the presence of silica, and then extracted. RESULTS AND DISCUSSION
Figure 1 shows a molecular weight profile obtained from an HRTLC spot following extraction and methylation. The two major components, m/z (1118.1)21 and (1131.7)21, support the expectation of disialyl GSLs and their mass difference indicates the ions to be doubly charged. The doublet is characteristic of sphingosine (C18) and eicosasphingosine (C20) heterogeneity located in the ceramide moiety and the profile suggests the structures to be isomeric with either GD1a
30
REINHOLD AND SHEELEY
FIG. 2. Disialoganglioside fragmentation by MS2; isolation and resonance ejection of the doubly charged ion, m/z 1131.721 from Fig. 1.
or GD1b. Several minor components were also resolved in this extract and their molecular-weight-related profiles indicate them to be GD2, GM3, GA1, GM2, and GD3. The amounts of these materials were insufficient for confirmation. Ion selection of the major component, (1131.7)21, was accomplished by scanning the amplitude of the fundamental RF voltage in a reverse-then-forward manner while applying a resonance signal which ejected all ions greater or less than m/z 1131.7 (61 Da). Fragmentation of the selected ion was induced by resonance excitation (a ‘‘tickle’’ voltage) which alters the ion trajectory from the center of the trap causing CID with the helium damping gas. The product ion spectrum is presented in Fig. 2. Most of the major ions were comparable to data obtained with a triple-quadrupole instrument, but significant qualitative and quantitative differences indicate the modes of excitation are clearly different. The ceramide- (m/z 604.5) and carbohydrate- (m/z 828.9) containing fragments were easily discernible, although the former was in lower abundance. Fragments for Neu5Ac (m/z 375.9/397.9), a nonreducing terminal Hex-HexN (m/z 486.1), and a tandem-linked Neu5Ac–Neu5Ac (m/z 737.3/759.4) were at comparable intensities. The remaining fragments can be identified as products of one or more lost termini and this spectrum clearly defines a GD1b sequence. An ion (m/z 847.6) strongly indicates that GD1b is not the
only isomer present. This fragment, (Neu5Ac-HexHexNAc z Na)1, argues for a small amount of GD1a, which was also present when the glycan itself was isolated and activated (see below, Fig. 4). However, no linkage information was detected. Selection of the ceramide fragment m/z 604.5 by again ejecting all ions greater and less than m/z 604.5 (by 61 Da) followed by CID with a tickle frequency provided the spectrum in Fig. 3. Methanol elimination (m/z 572.1) supports an original 3-hydroxy eicosasphingenine which again was suggested by the abundant ion at m/z 306.3, a ceramide product following N-acyl elimination. This major fragment strongly suggests the neutral loss N-acyl fatty acid moiety to be stearic acid by mass difference (see inset, Fig. 3). Isolation and activation of the carbohydrate-containing fragment, m/z 828.8, provided linkage information, in addition to the expected sequence and branching detail (Figs. 4 and 5). The selected doubly charged ion expelled a sodium ion upon activation (m/z 1634.9) and also showed losses of single and tandem-linked neuraminyl residues (m/z 641.121/1259.7 and 898.3, respectively). Losses of the two nonreducing terminal fragments, Hex-HexN (m/z 486.0) and the mono- and tandem neuraminyl residues (m/z 375.8/397.7 and 737.3/759.2), supported the proposed earlier GD1b structure (Fig. 2) with fragment m/z 847.6 suggesting a small amount of GD1a. An additional loss of the reduc-
MULTISTEP SPECTROMETRIC ANALYSIS OF GLYCOSPHINGOLIPIDS
31
FIG. 3. Ceramide characterization by MS3; isolation and resonance ejection of the fragment ion, m/z 604.5 from Fig. 3.
FIG. 4. Component carbohydrate characterization by MS4; isolation and resonance ejection of the fragment ion, m/z 828.821 from Fig. 3.
32
REINHOLD AND SHEELEY
FIG. 5.
Scale expansion of Fig. 4 (MS4) for detection of cross-ring fragments.
ing terminal glucosyl, residue with one (m/z 1055.4) or two (m/z 675.9) neuraminyl losses, and fragment m/z 898.3 supports the core sequence of this GSL, HexHexN-(HO)Hex-Hex-OH. Closer examination also indicates an ion 14 Da lower than these fragments (Fig. 5), again supporting a small amount of the GD1a isomer with the same sequence, but with modified Neu5Ac branching [HOHex-HexNAc-(HO)Hex-Hex (e.g., m/z 1040.1 and 884.3)]. Additional structural detail was apparent upon expansion of the low-mass end of the spectrum (Fig. 5) where a series of cross-ring cleavages (or their absence) were detected, suggesting the backbone linkage structure, Hex-3HexN-4(HO)Hex-4HexOH, m/z 574.0 and 1143.5, and the neuminyl linkage, m/z 545.3 (although this fragment would be isobaric for 237, 238, or 239 ketosidic linkages). Additional reducing-end fragments were observed and also losses of nonreducing termini which corroborate the overall glycan structure. Assuming the usual monomer motif for GSL and their anomeric configurations, the structure presented in Scheme 1 can be proposed for the major component. Mass spectral techniques using multiple low-energy collisions provided by Q1q2Q3 instrumentation have proven to be effective strategies for understanding carbohydrate structural detail, e.g., molecular topology (sequence and branching) from single and multiple glycosidic bond cleavage and linkage information from
cross-ring cleavage fragments pendent to nonreducing termini (3, 11, 12). They are particularly abundant in neutral oligomers where collision energy cannot be dissipated easily and are summarized in Scheme 2. Thus, from the nonreducing terminus, cross-linkage fragments were observed for the branched Neu5Ac(28)Neu5Ac linkage and are not expected for the back-
SCHEME 1. Summary of proposed ion fragments determined for the disialoganglioside GD1b from MSn spectra.
MULTISTEP SPECTROMETRIC ANALYSIS OF GLYCOSPHINGOLIPIDS
33
bohydrate sequencing by ITMS need to be clarified and further studied, this preliminary report has identified GSL linkage detail that was not observed with Q1q2Q3 instrumentation. ACKNOWLEDGMENTS
SCHEME 2. Cross-ring cleavage fragments observed here and in neutral oligosaccharides (11, 12) and detected in MS3 and MS4 spectra (Figs. 4 and 5). The masses identified are increments above the normal glycosidic fragment shown in parentheses in Scheme 1.
bone Hex(1-3)HexNAc linkage. Cross-ring fragments were also observed as consistent, albiet minor, fragments for the subsequent HexNAc(1-4)Hex and Hex(1-4)Hex, at m/z 574.0 and 1143.5 (Figs. 5 and 4, respectively). Unfortunately, these structural features are not uniformly detectable and collision spectra of many glycoconjugates are incompletely characterized. A primary constraint is that higher molecular weight samples show a general decrease in the extent of fragmentation with increasing size. Adding further complication is the fact that specific residues (HexNAc, NeuNAc) exhibit facile glycosidic rupture yielding product ions that dominate the spectra at the expense of less abundant, more informative linkage and branching fragments. This study had the expectation that an additional one or two dimensional collision experiments (MS3– 4) could isolate selected components of structure both smaller in size and possibly free from the dominant energy sinks of facile bond cleavage to reveal linkage and branching topology. Although many details of car-
We thank Bruce B. Reinhold for his careful review of the manuscript and offering important suggestions and the Department of Analytical Sciences, Glaxo Wellcome Inc., for use of the LCQ. The human brain disialoglycosphingoside sample was provided by H. C. Yohe, VA Medical and Regional Office Center (White River Junction, VT). This work has been supported in part by NIH Grant R01 GM54045 (V.N.R.).
REFERENCES 1. Fenn, J. B., Mann, M., Meng, C. K., Wong, S. F., and Whitehouse, C. M. (1989) Science 246, 64 –71. 2. Karas, M., and Hillenkamp, F. (1988) Anal. Chem. 60, 2299 – 2301. 3. Reinhold, B. B., Chan, S.-Y., Chan, S., and Reinhold, V. N. (1994) Org. Mass Spectrom. 29, 736 –746. 4. Stroud, M. R., Handa, K., Ito, K., Salyan, M. K., Fang, H., Levery, S. B., Hakomori, S., Reinhold, B. B., and Reinhold, V. N. (1995) Biochem. Biophys. Res. Commun. 209, 777–787. 5. Stroud, M. R., Salyan, M. E. K., Ito, K., Handa, K., Levery, S. B., Hakomori, S.-I., Reinhold, B. B., and Reinhold, V. N. (1996) Biochemistry 35, 758 –769. 6. Whitehouse, C. M., Dreyer, R. N., Yamashita, M., and Fenn, J. (1985) Anal. Chem. 57, 675– 679. 7. Schwartz, J. C., and Jardine, I. (1996) Methods Enzymol. 270, 522–586. 8. March, R. E. (1997) J. Mass Spectrom. 32, 351–369. 9. Jonscher, K. R., and Yates, J. R. (1997) Anal. Biochem. 244, 1–15. 10. Ciucanu, I., and Kerek, F. (1984) Carbohydr. Res. 131, 209 –217. 11. Reinhold, V. N., Reinhold, B. B., and Costello, C. E. (1995) Anal. Chem. 67, 1772–1784. 12. Reinhold, V. N., Reinhold, B. B., and Chan, S. (1996) Methods Enzymol. 270, 377– 402.