ANALYTICAL

BIOCHEMISTRY

192,181-l%

(1991)

Structural Classification of Carbohydrates in Glycoproteins by Mass Spectrometry and HighPerformance Anion-Exchange Chromatography’ John

R. Barr,*

Kalyan

R. Anumula,?

Michelle

Paul

B. Vettese,*

B. Taylor,7

and Steven

A. Carr*p2

Departments of *Physical and Structural Chemistry, and tMacromolecular Sciences, Smith Kline Beecham Pharmaceuticals, King of Prussia, Pennsylvania 19406

Received

July

5, 1990

A general strategy has been developed for determining the structural class (oligomannose, hybrid, complex), branching types (biantennary, triantennary, etc.), and molecular microheterogeneity of N-linked oligosaccharides at specific attachment sites in glycoproteins. This methodology combines mass spectrometry and high-performance anion-exchange chromatography with pulsed amperometric detection to take advantage of their high sensitivity and the capability for analysis of complex mixtures of oligosaccharides. Glycopeptides are identified and isolated by comparative HPLC mapping of proteolytic digests of the protein prior to, and after, enzymatic release of carbohydrates. Oligosaccharides are enzymatically released from each isolated glycopeptide, and the attachment site peptide is identified by fast atom bombardment mass spectrometry (FAB-MS) of the mixture. Part of each reaction mixture is then permethylated and analyzed by FABMS to identify the composition and molecular heterogeneity of the carbohydrate moiety. Fragment ions in the FAB mass spectra are useful for detecting specific structural features such as polylactosamine units and bisecting N-acetylhexosamine residues, and for locating inner-core deoxyhexose residues. Methylation analysis of these fractions provides the linkages of monomers. Based on the FAB-MS and methylation analysis data, the structural classes of carbohydrates at each attachment site can be proposed. The remaining portions of released carbohydrates from specific attachment sites are preparatively fractionated by high-performance anion-exchange chromatography, permethylated, and analyzed by FAB-MS. These analyses yield the charge

1 This work was supported Institutes of Health GM-39526 ’ To whom correspondence dressed. 0003-2697/91 Copyright All rights

in part by a grant from the National to S. A. Carr. and reprint requests should be ad-

$3.00 0 1991 by Academic Press, of reproduction in any form

state and composition of each peak in the chromatographic map, and provide semiquantitative information regarding the relative amounts of each molecular species. Analytically useful data may be obtained with as little as 10 pmol of derivatized carbohydrate, and fmol sensitivity has been achieved. The combined carbohydrate mapping and structural fingerprinting procedures are illustrated for a recombinant form of the CD4 receptor glycoprotein. 0 1991 Academic Press,

Inc.

In recent years, the structure and biological activities of glycoproteins have generated intense interest. Although glycosylation is perhaps the most common posttranslational modification of proteins, the biological significance of the carbohydrate moieties of many glycoproteins are not currently well defined. Oligosaccharides are known to affect physical properties of the glycoprotein such as solubility and stability, but they may also be more subtly involved in processes such as immunogenicity, molecular recognition, and systemic clearance (l-8). Glycoproteins are being studied as targets for the development of new, biologically active agents, and as therapeutic entities in the biopharmaceutical industry (1,2,9-18). The attachment site of Asn-linked oligosaccharides is generally present in the consensus sequence Asn-X-Ser/Thr (where X may be any amino acid except proline) (19). However, only selected consensus sequence sites are glycosylated. Furthermore, oligosaccharides at any given attachment site are generally a heterogeneous population of species. This heterogeneity could include structural class (complex, hybrid, or high mannose) or differences in molecular structure within the same structural class (including number, type, or linkage of sugar residues). Currently, the majority of structural data obtained on the oligosac181

Inc. reserved.

182

BARR ET AL.

charide moieties of glycoproteins has been on mixtures of carbohydrates derived from different attachment sites (2,20). Consequently, it has not been generally possible to correlate specific carbohydrate structures attached to specific regions of the folded, active protein with the activity of the molecule. Fast atom bombardment mass spectrometry (FABMS)3 in combination with other analytical techniques, such as Edman degradation, has been shown to be a powerful tool for characterizing post-translational modifications of proteins (21-23), including carbohydrates (24-33). In this report, we describe a new strategy employing high-performance anion-exchange chromatography with pulsed amperometric detection (HPAEPAD), methylation analysis, and mass spectrometry to define the structural class and microheterogeneity of N-linked oligosaccharides at specific attachment sites in glycoproteins. This procedure, in combination with the previously described mass spectrometry-based carbohydrate mapping strategy (24-26), allows the rapid determination of attachment sites, structural class (oligomannose, hybrid, complex), branching (biantennary, triantennary, etc.), and both types and amount of microheterogeneity of the carbohydrate moieties present at each specific N-linked attachment site present in a glycoprotein. Since this methodology can produce a great deal of information in a rapid fashion, it can be employed to characterize the carbohydrate moieties of a variety of recombinant glycoproteins and thus, evaluate the effects of cell type, growth conditions, and purification on the structural diversity of oligosaccharides at each attachment site of the glycoprotein. This is important since cell type, growth conditions (1,34), and purification (35) often have a profound affect on glycosylation and, therefore, can alter the physiological activity of the protein (1). The chemistry employed in the present strategy has been simplified and improved over the earlier report (26), resulting in greater reproducibility and two orders of magnitude improvement in sensitivity down to the low picomole range. Here we describe application of this optimized methodology to obtain structural data on the carbohydrate moieties at each attachment site of a soluble form of the CD4 antigen (sCD4) expressed in Chinese hamster ovary cells. MATERIALS

AND

METHODS

Preparation of glycopeptides. The sCD4 receptor glycoprotein was cloned, expressed, isolated, and purified in the same manner as described previously (14,33). 3 Abbreviations used: HPAE, high-performance anion-exchange chromatography; PAD, pulsed amperometric detection; HIV, human immunodeficiency virus; sCD4, soluble CD4 receptor; FAB-MS, fast atom bombardment mass spectrometry; RP, reversed phase; RCM, reduced and carboxymethylated, PNGase F, peptide:N-glycosidase F; Hex, hexose; HexNAc, N-acetylhexosamine; dHex, deoxyhexose; Fuc, fucose; DMSO, dimethyl sulfoxide; tPA, tissue plasminogen activator.

Reduction and carboxymethylation. sCD4 was reduced and carboxymethylated in a similar fashion as reported previously (23). Trypsin digestion. The reduced and carboxymethylated sCD4 sample (l-30 nmol) was dissolved in 0.250.5 ml of 50 mM ammonium bicarbonate buffer (pH adjusted to 8.2 with ammonium hydroxide) and treated, with a 1:lOO (w/w) enzyme:substrate ratio of trypsin (TPCK treated; Worthington) in 50 mM ammonium bicarbonate buffer (pH 8.2). The resulting solution was then maintained at 37’C under argon for 3 h and a second aliquot of trypsin was added in the pH 8.2 bicarbonate buffer (which afforded an enzyme:substrate ratio of 1:50). This mixture was then allowed to react for an additional 3 h. The reaction was quenched by the addition of acetic acid (final pH was 5-6) prior to analysis by reversed-phase HPLC. Glycosidase digestion with peptide:N-glycosidase F (27). An aliquot (0.1-10 nmol) of the reduced and alkylated sCD4 was dissolved with the aid of ultrasonication in 150 /*l of 100 mM ammonium bicarbonate buffer containing 0.005% EDTA (pH of buffer adjusted to 8.5 with NH,OH). To this solution was added 1 unit (in 4 ~1 of glycerol solution; a unit is defined as the amount of PNGase required to hydrolyze 1 nmol of fetuin glycopeptide per minute at 37’C, pH 8.6) of peptide:N-glycosidase F (Genzyme) and the resulting solution was shaken under argon for approximately 20 h. Small amounts of protein (100 pmol) only require 0.25 unit of PNGase. The reaction was quenched by the addition of 10 ~1 of 5% acetic acid. Aprotinin (0.01 pg in 1~1; freshly prepared), a tryptic inhibitor from bovine lung (Boehringer-Mannheim), was added to deglycosylation reactions involving enzymatic digests. Other enzymes such as Endo H may also be employed in these studies (24,25). HPLC. Reversed-phase HPLC separations were performed on a Beckman System Gold equipped with a 126 programmable solvent module and a 166 variable wavelength detector. Data collection and manipulation, in addition to system control, were accomplished on an IBM-PC AT with Beckman System Gold software. Fractionation of tryptic digests were performed on a fully endcapped Vydac Cl8 (25 cm X 4.6 mm) reversed-phase HPLC column. Solvent A was 0.1% aqueous trifluroacetic acid (v/v) and solvent B was 90% acetonitrile/lO% water and 0.1% trifluroacetic acid (v/v). Typically, gradients started at O-5% B and proceed to 80% B at 0.751% B/min. Permethylationof otigosaccharides. Free oligosaccharides may be derivatized in the presence of protein with a system similar to Ciucanu and Kerek’s (36). The permethylated oligosaccharide derivatives may then be easily purified from peptides and salts via a simple water/ chloroform partition. Oligosaccharides that contain

SPECTROMETRIC

STRUCTURAL

ANALYSIS

sialic acid often fail to completely form the methyl ester of the acid moiety. Treatment with diazomethane usually completes this transformation and yields the desired permethylated species. An aliquot (100 pmol-10 nmol) of PNGase treated glycopeptide was placed in a l-ml thick-walled glass reaction vial (which had been silylated with CH,Cl,Si) and concentrated to dryness under reduced pressure. The resulting residue was then redissolved in 100 ~1 of DMSO (Aldrich) and 8.0 mg (200 pmol) of powdered sodium hydroxide and a magnetic stir bar were added to the system. Argon was bubbled through the solution during the addition of the reagents. Methyl iodide (Fluka; 25 ~1,400 pmol) was then added to the reaction mixture and the resulting solution was again flushed briefly with argon and allowed to stir for 20-30 min. The reaction was then quenched by the addition of 200 ~1 of 30% acetic acid and extracted with three portions of 200 ,ul chloroform. The organic extracts were combined, washed with two portions of 200 ~1 of 30% acetic acid, and concentrated to dryness under reduced pressure. The methylated oligosaccharide was then redissolved in 200 ~1 of methylene chloride and treated with diazomethane (-6 mM in 2:l ether:ethanol) dropwise until a yellow color persisted. The resulting solution was then allowed to stir under ambient conditions for 15 min. A second aliquot (2-3 drops) of the diazomethane solution was then added and the reaction mixture was allowed to stir for an additional lo-15 min. This solution was then concentrated to dryness under reduced pressure. High-performance anion-exchange chromatography. Oligosaccharides, released from glycopeptides, may be fractionated by HPAE. Complex carbohydrates may be separated according to charge state or other microheterogeneity (37,38). In general, larger oligosaccharides are retained longer than smaller oligosaccharides of the same charge state. Fucosylated oligosaccharides, however, tend to elute faster than the similar (but smaller) nonfucosylated sugars (37). This separation allows the microheterogeneity at each attachment site to be defined and quantitated. Glycopeptide containing RP-HPLC fractions (l-10 nmol) were treated with PNGase F (vide supra) to release the carbohydrate portion and the reaction mixture was concentrated to dryness under reduced pressure. The resulting residue was then dissolved in water and applied to a 2 ml AG-50W-2X (H+ form; ZOO-400 mesh, 0.6 meq of salt/ml of wet resin) column using a fivefold excess of resin to expected salt to desalt the fraction. The oligosaccharides were eluted with 5 column volumes of water. Fractionation of the mixture was accomplished by high-performance anion-exchange chromatography on a Dionex Bio-LC. This system consists of a gradient pump, a gold working electrode for pulsed amperometric detection, and an advanced computer inter-

OF GLYCOPROTEIN

CARBOHYDRATES

183

face with A1420 software. An IBM compatible PC was employed for data collection and handling. A polymeric pellicular anion-exchange column (4 x 250 mm HIPCAS6 CarboPac PA-l analytical column from Dionex) equipped with a CarboPac guard column was used in all experiments. Samples were injected using a Spectra Physics SP8780 autosampler equipped with a Tefzel rotor seal in a Rheodyne injection valve. Chromatographic separations of the oligosaccharides were performed at 1.0 ml/min using the following gradient; isocratic at 175 mM NaOH and 44 mM NaOAC for 2 min followed by a linear gradient to 200 mM NaOAc and 175 mM NaOH at 55 min. The pH of the sodium acetate solution was 5.5. Detection was achieved with a pulsed amperometric electrochemical detector with a gold working electrode at the following potentials: E, = 0.01 V (tl = 0.3 s), E, = 0.7 V (t2 = 0.12 s), E, = -0.3 V (t3 = 0.3 s). A response time of 3 s was used for PAD 2. The solvents and gradient employed in this study were fairly standard. Under similar conditions some epimerization of the reducing-end sugar and /3-elimination of 3-linked substituents on the reducing end HexNAc have been previously observed (3839). In addition, O-acetate groups tend to degrade when subjected to the alkaline conditions employed for the chromatography (39). It has been reported that some of these types of degradation can be minimized through the use of milder alkaline conditions such as 30 mM NaOH. However, in the present study, no structural modification of oligosaccharides derived from sCD4 or standards was observed. The carbohydrate-containing fractions were then generally desalted on AG 5OW-2X (in a similar manner as above), concentrated to dryness under reduced pressure, and subjected to treatment with DMSO/NaOH/ CH,I (see permethylation section above). The reaction mixture was then acidified with 30% acetic acid and the permethylated oligosaccharides were purified from salts with a water-chloroform partition. Mass spectrometry. Fast atom bombardment mass spectra were recorded on either the first double focusing portion of a VG ZAB SE-4F tandem magnetic deflection mass spectrometer which employs an accelerating voltage of 10 kV and a mass range of 12,500 or on a VG ZAB-HF magnetic deflection mass spectrum (accelerating voltage, 8 kV; mass range, 3000). The VG ZAB SE4F was equipped with a flow FAB ion source and a Cs ion gun operated at 35 kV and with an emission current of 2-4 PA. The VG ZAB-HF was also equipped with a standard flow FAB ion source, but employed an ION Tech FAB gun that was operated at 8 kV and utilized a discharge current of 1 mA. All data were acquired and processed on a VG 11-2505 data system. Typically, 0.05 to 1 nmol of a proteolytic digest or glycopeptide-containing HPLC fraction was analyzed by FAB-MS using a matrix of thioglycerol (Sigma). Samples were gener-

184

BARR

t

t +

PEPTIDES

PEPTIDES

GLYCOP+EPTlDES -hl-

FREE CARBOHYDRATES

t

(CHO)

MIALYTICAL $ I I

MAP OF PROTEIN (EXCLUDING REGIONS AROUND GLYCOSYLATION SITES)

HPLC

J 1

OBSERVEEXPECTED GLYCOSYLATION-SITE PEPTIDE SHIFTED UPWARDS BY ONE MASS UNIT (N=114; D=115)

COMPARE HPLC PROFILES T O IDENTIFY PUTATIVE GLYCOPEPTIDESAND DEGLYCOSYLATED PEPTIDES

PREP HPLC 1 ISOLATE GLYCOPEPTIDES

FIG. 1. gosaccharides

Strategy for identifying in glycoproteins.

attachment

sites of Asn-linked

oli-

ally analyzed in two experiments that employed overlapping mass ranges. The higher mass range (ca. 41001850; 100 s/decade) was examined on the VG ZAB SE4F at low resolution (R - 800), while the lower mass range (ca. 2250-100; 25 s/decade) was generally examined on the VG ZAB-HF at a resolution of 2000. Full spectra (ca. m/z 3100-300; 40 s/decade) were recorded in a single experiment on the ZAB SE-4F with the Cs gun operating at 20 KV. The FAB mass spectra of the purified permethylated oligosaccharides were recorded in a similar fashion as described above with the exception that a matrix of 1:l thioglycerol:m-nitrobenzyl alcohol (Aldrich) plus 1% trifluroacetic acid (Sigma, HPLC grade) was employed. This matrix appears to afford the greatest sensitivity for the permethylated oligosaccharides. Several scans were often required before ions from the purified permethylated oligosaccharides were observed. Generally, 100 pmol-1 nmol of permethylated oligosaccharide was employed [lower amounts (1 pmol) may sometimes be used]. Oligosaccharides lacking sialic acid residues generally can be analyzed using lo-fold less sample. RESULTS

AND

DISCUSSION

The methodology begins with carbohydrate mapping of the glycoprotein to identify and isolate glycopeptides derived from proteolytic digestion (Fig. 1). Ions for gly-

ET

AL.

copeptides are often not observed in fast atom bombardment mass spectrometry. It is generally believed that the hydrophilic nature of the carbohydrate moiety lowers the surface activity in the matrix relative to peptides. Reduced and carboxymethylated glycoproteins are digested with trypsin (or another suitable enzyme), and an endoglycosidase (27). Generally, PNGase F has been employed to release oligosaccharide moieties because of its broad specificity (hydrolyzes N-linked bi-, tri-, and tetraantennary oligosaccharides, in addition to polysialyl complex carbohydrates, and sulfate-containing oligosaccharides), and PNGase F converts the glycosylated Asn to Asp which aids in the determination of the attachment site by mass spectrometry since Asp weighs one dalton more than Asn (24,25,27). Glycopeptide-containing fractions are identified by comparing the HPLC profiles of proteolytic digests of reduced and carboxymethylated glycoproteins prior to, and after, treatment with PNGase F. This is illustrated by the tryptic digest of reduced and alkylated sCD4 in Fig. 2. Subsequently, the putative glycopeptide-containing fractions are treated with PNGase F, and the products are analyzed by FAB-MS (Fig. 3). In the case of sCD4, signals were observed in the FAB mass spectra that corresponded to the expected (M + H)+ for the tryptic peptides Ly~‘~~-Lys~~‘, Leu253-L Ys27g,Ala2g4Lys312,and Asn3”-Lys312 in which Asn271and Asn300had been converted to Asp upon release of the carbohydrate units. These experiments demonstrated that both potential glycosylation sites in sCD4 were utilized (23). Signals corresponding to the Asn-containing tryptic peptides were not detected prior to treatment with the glycosidase indicating that both sites were essentially fully glycosylated. Release of the carbohydrate moiety of a glycoprotein with PNGase F can be achieved before or after proteolytic cleavage. However, in cases that proteolytic digestion generates glycosylation sites on the N-terminal amino acid of a peptide, incomplete release of the carbohydrate unit is often observed with PNGase F. sCD4 contains an expected glycosylation site on the N-terminal amino acid of a tryptic fragment (Asn300-Lys312)and inefficient release of the oligosaccharide was observed for this glycopeptide. Therefore, complete hydrolysis of the oligosaccharide from this site was obtained by subjecting RCM-sCD4 to PNGase F treatment prior to tryptic digestion. A portion of the PNGase F reaction mixture for each glycopeptide is then permethylated and extracted (Fig. 3). With the DMSO/NaOH/CH,I methylation procedure, over and under methylation was not found. An aliquot of the chloroform extract from this reaction is analyzed by FAB-MS to identify the composition of the oligosaccharides at each attachment site (see Fig. 4). Due to the lipophilic nature of permethylated oligosaccharides, the protected derivatives can be easily purified from peptides and salts through a simple water/chloro-

SPECTROMETRIC

E: :: d

35

40I

45I

ANALYSIS

OF

GLYCOPROTEIN

185

CARBOHYDRATES

through the comparison of the intensities of the (M + H)+ signals for each permethylated carbohydrate. The possible carbohydrate compositions corresponding to each of the observed (M + H)+ signals are generated by computer, typically employing windows of O-10 units of hexose (Hex, 204.2 Da) and N-acetylhexosamine (HexNAc, 245.3 Da) and O-5 units of sialic acid (NeuAc, 361.4 Da) and deoxyhexose (dHex, 174.2 Da). Molecular weights are calculated by summing the various in-chain masses noted above with the masses of the methyl end groups (CH, + OCH,, 46.1 Da). In the case of sCD4, this analysis suggested that the carbohydrate moieties for both Asn271 and Asn300 were composed of oligosaccharides with the general formula NeuAc,Hex,HexNAc,d Hex,, (where x = O-2, y = 0,l). At each attachment site, the most abundant glycoforms were those containing only one sialic acid group. The FAB-MS analysis also suggested that Asn271 contained approximately equal amounts of glycoforms that included and lacked the dHex residue (Fig. 4). In contrast, the FAB-MS data suggested that the’ oligosaccharide unit of Asn300 contains a higher percentage of glycoforms that include a dHex group than the oligosaccharides at Asn271 (data not shown). Significant structural information can be derived from careful examination of the fragment ions in the FAB mass spectra of the permethylated glycoforms. Permethylation of an oligosaccharide enhances the sen-

J

0 cl ; .j, D 30,

STRUCTURAL

501

Time

60I

551

65I

I JJ 70

75 810

GLYCOPEPTIDE-CONTAINING FRACTIONS I

(min)

(Top) HPLC trace of 2 nmol of a tryptic digest of reduced FIG. 2. and carboxymethylated (RCM) soluble CD4 for comparison with (bottom) the HPLC trace of the trypsin digest of PNGase F-treated RCM-soluble CD4. Only the regions ofthe chromatograms that exhibited peak shifts are shown. Detection was accomplished by uv at 215 nm. Mobile phases were as described in the text. Gradient: 5 min hold at 5% B; 5% B to 18% B in 18 min; 18% B to 45% B in 90 min; 45% B to 90% B in 32 min. Peaks containing glycopeptides (top) and former glycopeptides (bottom) are filled in.

form extraction. Carbohydrate units of glycoproteins at a specific attachment site are generally composed of a collection of glycoforms, thus, multiple molecular ions are often observed. In general, FAB mass spectra of permethylated oligosaccharides are dominated by pseudomolecular ions and useful fragmentation. These experiments are typically done at reduced resolution on the mass spectrometer (500-800) in order to maximize sensitivity. The mass values obtained for the permethylated oligosaccharides are therefore chemical average rather than monoisotopic. The FAB mass spectrum of the family of permethylated oligosaccharides formerly attached to Asn2’l is shown in Fig. 4. An estimate of the relative amounts of each glycoform can be obtained

FASMS -

PEPTIDES FABMS + OLIGOSACCHARIDES FROM SPECIFIC ATTACHMENT SITES

OBSERVE GLYCOPEPTIDE (IN FAVORABLE CASES)

IDENTIFY FORMER GLYCOSYLATIONSITE PEPTIDE

I P;R,Fl;D ’ OL~GOSACCHARIDES

LINKAGE OF MONOMERS

4

IDENTIFY COMPOSITIONS AND MICROHETEROGENEITY OF CARBOHYDRATES

1 DESALT 2 PERMETHYLATE 3 FABMS

\ v PROPOSE STRUCTURAL CLASSES OF CARBOHYDRATES AT SPECIFIC ATTACHMENT SITES

FIG. neity

I CONFIRM CHARGE STATES; DEFINE COMPOSITION AND HETEROGENEITY OF OLIGOSACCHARIDES IN FRACTION

Strategy for identifying structural class and microheteroge3. of oligosaccharides at specific attachment sites.

186

BARR

loo.

Gall-

344 376

/

a g F 3 E

4GlcNAcl

ET

AL.

-

2Manl

\ @an1

NeuAc n >

100

Galld4GlcNAcl

-W PManl

2410 2133

/’

-4GlcNAcl ndl-2 x=0-1

Fuc, 11 ts +4GlcNAc

2584

NeuAcj Fuco

NeuAcl Fuel

NeuAc2 Fuco

50

2772

I

NeuAc2 Fuq 2946

m/r FIG. 4. (A) FAB mass spectrum of ca. 200 pmol of the permethylated carbohydrates released digestion of glycopeptide fractions 69-72 from the tryptic digest of reduced and carboxymethylated shown (biantennary oligosaccharides with varying numbers of NeuAc, some with fucose attached from the combination of the FAB-MS data and methylation analysis data (see Fig. 3 and text).

sitivity for detection which aids in obtaining molecular weight information. Furthermore, the FAB mass spectra of these derivatives exhibit abundant fragments that yield considerable sequence information. The major fragmentation pathway generally observed for permethylated (or peracetylated) oligosaccharides involves glycosidic bond cleavage resulting in the formation of an oxonium ion and charge retention on this nonreducing terminal fragment (A-type cleavage) (28). Fragmentation of permethylated oligosaccharides usually occurs most readily (sometimes exclusively) at the reducing end of hexosamine residues and neuraminic acid. As a consequence, the FAB mass spectra of permethylated derivatives do not usually provide direct structural information about the Man,GlcNAc core. The low probability of cleavage adjacent to neutral sugars can also occasionally make it difficult to distinguish positional isomers of certain substructures located at the nonreducing chain termini. For example, NeuAc-GalGlcNac- cannot be distinguished from Gal-GlcNAc (NeuAc)- by MS alone. When we are required to answer

from Asn”’ attachment site by PNGase F soluble CD4 (Fig. 2 (top)). The structures to the reducing end GlcNAc) are derived

this question we treat the glycopeptides with neuraminidases from Vibrio cholerae versus neuraminidase from arthrobacter urifaciens and analyze the resulting carbohydrates (following release from the peptide) by HPAEPAD and MS (following permethylation). The former enzyme cleaves NeuAca2 + 6GlcNAc only very slowly and can therefore be used to distinguish these positional isomers (40). The FAB-MS of the glycoforms attached to Asn”l (Fig. 4) contain several structurally informative fragment ions that were very similar to the spectrum of the standard oligosaccharide shown in Fig. 5. The ion at m/z 825 corresponds to permethylated NeuAc-Hex-HexNAc+ (A-type cleavage). Further loss of methanol from the 825 ion then leads to the species indicated by the ion at mlz 793. Dell has suggested that the ion at ml2 793 indicates that the Hex residue is linked through the O-4 position of the HexNAc (28). This linkage was further supported by methylation analysis (23). The fragment ion at m/z 464 is Hex-HexNAc+ and is presumably derived from the A-type cleavage with the charge retained

SPECTROMETRIC

STRUCTURAL

ANALYSIS

OF

GLYCOPROTEIN

187

CARBOHYDRATES

m’25 3Gall-w

2Menl

4GlcNAcl

4GlcNAc NeuAc

2 --3Gall-,

4GlcNAcl

-

2Manl

500 Fmtomoln on Robe

NW 2n2

2495

925

I

‘?

2495

FIG. 5. FAB mass spectrum of 10 pmol of (NeuAc-Gal-GlcNAc-Man&-Man-GlcNAc-GlcNAc. The spectrum was recorded between 3100 and 320 (at 40 s/decade) employing a Cs ion gun operated at 20 kV. (Insert) FAB-MS spectrum of 500 fmol of (NeuAc-Gal-GlcNAc-Man),Man-GlcNAc-GlcNAc. The spectrum was recorded on a VG ZAB SE-4F. Only the molecular ion region is shown. The spectrum was recorded from 3100 to 615 (40 s/decade) employing a Cs ion gun operated at 35 KV.

on a nonreducing terminus that lacks sialic acid. The peaks at m/z 376 and 344 are ions resulting from the A-type cleavage of the sialic acid residues. Loss of the reducing terminal HexNAc or HexNAc-dHex residue from the (M + H)+ species at m/z 2410 or 2584, respectively, yields the oxionium ion at m/z 2133. The absence of a signal at m/z 2307 [corresponding to the loss of the permethylated reducing terminal HexNAc from the species NeuAc,HexNAc,Hex,dHex, (2584)] indicates that the dHex residue was attached to the HexNAc at the reducing end of the oligosaccharide. Ions present in the region m/z 1150-1300 in Fig. 4 do not appear to be carbohydrate related and are generally not observed. Strong FAB mass spectra can be obtained at the low picomole level with purified, permethylated oligosaccharides. The FAB mass spectrum obtained for 10 pmol of the commercially available standard biantennary oligosaccharide (NeuAc-Gal-GlcNAc-Man),-Man-GlcNAc -GlcNAc following permethylation is shown in Fig. 5. In

addition to the fragment ions described above, an Atype fragment was observed at m/z 2495 which corresponds to the loss of the reducing terminal GlcNAc from the (M + H)+. Fragment ions for permethylated NeuAc and (NeuAc-MeOH) are also present at m/z 376 and 344, respectively, but are difficult to distinguish because of background matrix ions. Molecular weight information alone can be obtained at the 500-fmol level with >lO:l signal-to-chemical noise (see Fig. 5). The remaining portion (oligosaccharides released from 10 nmol of glycopeptide) of the permethylated material was then used for methylation analysis by GCMS (23) to determine the linkage of monomers (see Fig. 3). The molecular weight and fragmentation information (for Asnz71, see Fig. 4) in combination with the methylation analysis allowed the structural classes of carbohydrate at each attachment site to be proposed. These data indicated that the major carbohydrates at both attachment sites in sCD4 (Asn*‘l and Asn300) were

188

BARR

composed of biantennary complex carbohydrates heterogenous in their sialic acid and fucose content (see structure, Fig. 4). By methylation analysis, the NeuAc was found to be linked entirely 2,3 to Gal since 2,6linked Gal was not observed. In addition, the fucose was indicated to be linked entirely l-6 to GlcNAc by the presence of a 1,4,6-linked GlcNAc. Trace levels of 1,2,4linked mannose and terminal mannose residues suggests minor amounts of triantennary and oligomannose structures, respectively. The presence of trace amounts of triantennary complex oligosaccharides at each attachment site were also indicated by HPAE-PAD chromatography (vide infra). In a previous study (26), the released carbohydrate moieties of tissue plasminogen activator (tPA) were peracetylated and permethylated, and analyzed by FAB-MS. The FAB mass spectra of the peracetylated carbohydrates tended to show a great deal of under-acetylation and were generally more difficult to interpret. Most importantly, peracetylation was used in these studies to provide the organic extractable derivative of the carbohydrate which was subsequently permethylated. Underacetylation therefore, probably resulted in poor extraction efficiency in the organic/aqueous partition and in a lowering of the yield. Employing standard mixtures of oligosaccharides and protein hydrolysate, we have compared the present procedure of direct methylation with NaOH/CH,I/DMSO to acetylation followed by permethylation. This study indicated that direct permethylation of oligosaccharides in the presence of peptides yields significantly higher yields (as determined by the intensity of product-related signals in the resulting FAB mass spectra) of the fully derivatized oligosaccharide than samples that were peracetylated, extracted and permethylated. Thus, permethylation is now exploited for both purification of the oligosaccharide from peptides and salts (via extraction) and structural information (see Fig. 3). The specific permethylation method employed is also critical to achieving analytically useful results. In the present study, the procedure of Ciucanu and Kerek (36) was employed to methylate all carbohydrates. We have compared this method to methylation with potassium tert-butoxidel DMSO (26,41-43), and the Hakamori reagent (44). The procedure of Ciucanu and Kerek utilizing NaOH/ DMSO/CH,I was found to be cleaner, more rapid, and methylated the carbohydrates to a greater extent than either of the other two reagents. The final stage of the strategy involves preparative fractionation of the released oligosaccharides from each glycopeptide by HPAE with pulsed amperometric detection (PAD) and analysis of the purified oligosaccharides by FAB-MS following permethylation (see Fig. 3). The HPAE-PAD allows a high-resolution chromatographic fingerprint of the types and relative amounts of each glycoform to be established for each attachment

ET

AL.

ASP

B

OllgoaaccharMer

C

(PNGase

A

F Rdeased)

E DI I

!

K n

AWP Oligoseccharides (PNGase F Released)

d

0

32

15 Retention

Time

45

(Min)

FIG. 6. Preparative fractionation of ca. 10 nmol

of PNGase F-released oligosaccharides from Asnz71 (top) and A.&@’ (bottom) of soluble CD4 using HPAE with pulsed amperometric detection (see text for conditions). Within each charge group the fucosylated oligosaccharide elutes earlier than the nonfucosylated analogue.

site. The retention of an oligosaccharide on HPAE chromatography is generally influenced by the size and charge state of the carbohydrate. Subtle differences such as branching and linkage can also be detected, however. The assignment for each individual peak in the HPAE chromatogram is usually by comparison of retention time to known standards or by methylation analysis. The procedure reported here utilizes FAB mass spectrometry to rapidly define the chemical nature (in terms of charge state and composition) of each oligosaccharide containing fraction. The released carbohydrate mixtures from each attachment site of sCD4 were desalted on Dowex 5OW-2X and preparatively fractionated by high-performance anion-exchange chromatography with pulsed amperometric detection (Fig. 6). Recovery from the Dowex column was tested for a standard biantennary carbohydrate (NeuAc2-GGall-4GlcNAc-1-BMan),Manl-4 GlcNAcl-4GlcNAc and was found to be 85%. Desalting fractions prior to HPAE was found to enhance the chromatography. The chromatograms of the glycoforms derived from Asn271 and Asn30” are shown in Fig. 6, top and bottom, respectively. Each fraction from HPAE was

SPECTROMETRIC

STRUCTURAL

ANALYSIS

OF

TABLE

GLYCOPROTEIN

1

Calculated Masses for Molecular and Fragment Ions for Permethylated Oligosaccharides from Asn271and Fractionated by HPAE-PAD” Fqx 4GlcNAcl+2Manl-w6 \ Gail

-4GkNAcl

-2Manl

189

CARBOHYDRATES

/

-3

ManI

-w

4GkNAcl n=O-2 x=0-1

+

Derived

16 4GkNAc

Fragments Fraction A B C D E

NeuAc 0 0 1 1 2 2

Fuc

MH+

MH+-GlcNAc

1 0 1 0 1 0

2223 2049 2585 2411 2946 2772

MH+-Fuc-GlcNAc -

2133 2495

2133 2495 -

m See Fig. 6, top.

then desalted, permethylated and subjected to FAB-MS analysis (see Fig. 3) to confirm the charge state and define the compositions (Table 1). Fraction A from HPAE was found to contain neutral species (no sialic acid) both with and without fucose. Oligosaccharides with one sialic acid residue elute next (B and C, Fig. 6), followed by two and three sialic acid residues (D and E, and F and G, respectively, Fig. 6). Fucosylated oligosaccharides elute before nonfucosylated oligosaccharides of the same charge state. Triantennary oligosaccharides were found in low abundance by HPAE-PAD chromatography and methylation analysis, but were not observed in FAB-MS. The HPAE-PAD chromatogram confirmed the relative abundance of glycoforms observed in the FAB mass spectra. Species containing one sialic acid residue were the most abundant at both Asnz71 and Asn300.Approximately equal amounts of glycoforms derived from As#‘l contain or exclude fucose, while Asn300 was found to be fucosylated to a much greater extent, which confirmed the previously obtained FAB-MS data (discussed above). Because the molar response of a given carbohydrate in HPAE-PAD chromatography is affected by the degree of sialylation (38), more rigorous quantitation of the putative oligosaccharides may be obtained by desialylation of the putative glycopeptide pools prior to endoglycosidase release of carbohydrates and analysis by HPAE-PAD (45). In addition to the information on the composition of each fraction, the FAB mass spectra of each isolated and purified glycoform provides further insights into their structure. For example, the spectra of the permethylated oligosaccharides derived from fractions D and E (see Fig. 7 in addition to Table 1) contain a fragment ion at m/z 2495. This represents the A-type cleav-

age between the GlcNAc units at the reducing end, and indicates loss of a GlcNAc from the glycoform in fraction E [(NeuAc-Gal-GlcNAc-Man),-Man-GlcNAcGlcNAc] and Fuc-GlcNAc from the glycoform in fraction D [ (NeuAc-Gal-GlcNAc-Man),-Man-GlcNAc-(Fuc)-GlcNAc]. No peak is observed at m/z 2669 which would correspond to the loss of the reducing terminal GlcNAc from the glycoform in fraction D. The fucose residue must therefore reside on the GlcNAc at the reducing terminus of the oligosaccharide (methylation analysis suggests a l-6 linkage). Additionally, the permethylated oligosaccharides derived from fractions C [NeuAc,(Gal-GlcNAc-Man),-Man-GlcNAcGlcNAc] and B [NeuAc,(Gal-GlcNAc-Man),-ManGlcNAc-(Fuc)-GlcNAc] both showed a fragment ion in their FAB mass spectra at m/z 2133. This again corresponds to the loss of the permethylated GlcNAc from the reducing end of the oligosaccharide from fraction C and GlcNAc-Fuc from the permethylated glycoform in fraction B. No peak was observed at m/z 2307 (2584, permethylated reducing end GlcNAc) in the spectrum of the permethylated oligosaccharide from fraction B, which again indicates that the fucose on this oligosaccharide is located exclusively on the GlcNAc at the reducing terminus of the complex carbohydrate. In conclusion, this general strategy utilizes the high sensitivity and ability of mass spectrometry and HPAEPAD chromatography to analyze complex mixtures of oligosaccharides derived from specific attachment sites of glycoproteins. This methodology first identifies the attachment sites for N-linked oligosaccharides and isolates glycopeptides by carbohydrate mapping (24,25). The carbohydrate moieties of each glycopeptide are then released with PNGase F and a portion is methyl-

190

BARR

ET

AL.

Fuc, NeuAc

2 -3Galld

NeuAc

2

~GICNAC~

--3Gall-,

4GlcNAcl

-

2Manl

+: 4GlcNAc

+

2Mant MN’=

2646

2495 2231 I

2349

I

m/z

27 72

NeuAc

2 --3Gall+

4GlcNAcl

+

2495

2Manl L3

NeuAc

2 +3Gall--

4GlcNAcl

+

2Manl

6Manl MN’

2495

FIG. 7. FAB mass spectrum of (top) peak fucosylated (top) and nonfucosylated (bottom) terminal GlcNAc.

eluting in HPAE analogues contain

+4GltNA~l

--L

4GlcNAc

/ I 2772

2794

chromatogram (Fig. 6, top) at 20 min and (bottom) 21.5 min. Both the a fragment ion at 2495, indicating that the fucose residue is attached to the

SPECTROMETRIC

STRUCTURAL

ANALYSIS

ated and subjected to FAB-MS analysis. The FAB-MS data suggests the composition (in terms of hexose, deoxyhexose, N-acetylhexosamine, and sialic acid) and microheterogeneity of the oligosaccharides. Additionally, fragment ions yield a great deal of structural information such as branching and location of inner core deoxyhexose units. A second portion of the released carbohydrate is then subjected to methylation analysis which provides the linkage of monomers. FAB-MS in combination with data from the methylation analysis allows the structural class of oligosaccharides at each attachment site to be proposed. The remaining portion of the released carbohydrate is then preparatively fractionated by HPAE-PAD, permethylated, and analyzed by FAB-MS. These combined techniques define the charge state and composition of each peak in the HPAE chromatogram and indicate the relative amounts of each glycoform. Combining these complimentary techniques allows the rapid determination of structural class, branching type, and microheterogeneity of the major oligosaccharides at each specific attachment site in a glycoprotein. As noted here, minor glycoforms present at only a few percent of the major forms may escape detection by this methodology. However the major glycoforms identified in sCD4 by the procedures presented here (23) have subsequently been shown to be correct using more conventional procedures in which oligosaccharides released by hydrazinolysis and labeled with NaB3H, were analyzed by paper electrophoresis, lectin affinity and size exclusion chromatography, enzymatic microsequencing, and methylation analysis, thus confirming the general validity of the present approach (46). As of yet, we have only applied the methodology to glycoproteins containing relatively few glycosylation sites (~5). The applicability of the method to more complex glycoproteins such as gp120, the surface glycoprotein of the HIV virus, is currently under investigation. Recent advances in mass spectrometry, such as electrospray MS (47), should also be helpful for the observation of ions related to glycopeptides (48).

OF

GLYCOPROTEIN

191

CARBOHYDRATES

7. Hansen, L., Blue, Y., Barone, K., Collen, (1988) J. Biol. Chem. 263,15,713-l&719.

D., and Larsen,

G. R.

8. Galili, U., Clark, M. R., Shohet, S. B., Buehler, J., and Machar, B. A. (1987) Proc. Natl. Acad. Sci. USA 84, 1369-1373. 9. Kagawa, Y., Takasaki, S., Utsumi, chibe, N., and Kobata, A. (1988) 17,517. 10. Takeuchi, Kochibe,

M., Takasaki, N., and Kobata,

J., Hosoi, K., Shimizu, J. Biol. Chem. 263,

H., Ko17,508-

S., Miyazaki, H., Kato, T., Hoshi, S., A. (1988) J. Biol. Chem. 263, 3657-

3663. 11. Tsuda, E., Goto, M., Murkami, A., Akai, K., Ueda, M., Kawanishi, G., Takahashi, N., Sasaki, R., Chiba, H., Ishihara, H., Mori, M., Tejima, S., Endo, S., and Arata, Y. (1988) Biochemistry 27,5646-

5654. 12. Pohl, G., Kallstrom, H. (1984)

M., Bergsdorf, 23,3701-3707.

Biochemistry

N., Wallen,

13. Smith, D. H., Byrn, R. A., Marsters, man, J. E., and Capon, D. J. (1987)

P., and Jornvall,

S. A., Gregory, T., GroopScience 238,1704-1707.

14. Deen, K. C., McDougal, J. S., Inacker, R., Folena-Wasserman, Arthos, J., Rosenberg, J., Maddon, P. J., Axel, R., and R. W. (1988) Nature (London,J 33 1,82-84.

G., Sweet,

15. Fisher, R. A., Bertonis, J. M., Meier, W., Johnson, V. A., Costopoulos, D. S., Liu, T., Tizard, R., Walker, B. D., Hirsch, M. S., Schooley, R. T., and Flavell, R. A. (1988) Nature (London) 331,

76-78. 16. Hussey, R. E., Richardson, N. E., Kowalski, M., Brown, N. R., Chang, H.-C., Sihciano, R. F., Dorfman, T., Walker, B., Sodroski, J., and Reinherz, E. L. (1988) Nature fLondon) 331, 78-81. 17. Traunecker, A., Luke, (London) 33 1,84-86.

W.,

and

Karjalainen,

18. Berger, E., Fuerst, T. R., and Moss, Sci. USA 85,2357-2361. 19. Struck, D. K., Glycoproteins Plenum, New 20. Welply, J. A.

K. (1988)

B. (1988)

Proc.

Nature

Natl.

Acad.

and Lennarz, W. J. (1980) in The Biochemistry of and Proteoglycans (Lennarz, W. J., Ed.) pp. 35-83, York. (1989) Trends Biotechnol. 7, 5-10.

21. Carr, S. A., and Biemann, K. (1984) in Methods in Enzymology, (Estabrook, R. W., and Pullman, M. E.), Vol. 10, Part A, pp. 2958, Academic Press, Orlando, FL. 22. Biemann,

K., and Scoble,

H. A. (1987)

Science

23. Carr, S. A., Hemling, M. E., Folena-Wasserman, Anumula, K., Barr, J. R., Huddleston, M. (1989) J. Biol. Chem. 264, 21,286-21,295. 24. Carr,

S. A., and Roberts,

G. D. (1986)

Anal.

237,992-998. J.,

G., Sweet, R. W., and Taylor, P.

157,396-

Biochem.

406. ACKNOWLEDGMENTS We thank cussions.

Michael

25. Carr, Huddleston

and Dr. Mark

Bean

for helpful

dis-

S. A., and Roberts, quence Analysis (Walsh, ton, NJ.

G. D. (1987) in Methods K. A., Ed.), pp. 423-436,

26. Carr, S. A., Roberts, G. D., Jurewicz, Biochemie 70, 1445-1454.

REFERENCES 1. Parekh, R. B., Dwek, R. A., Edge, C. J., and Rademacher, (1989) TIBTECH 7, 117-122. 2. Rademacher, T. W., Parekh, R. B., and Dwek, R. A. (1988) Rev. Biochem. 67. 785-838. 3. Baenziger, J. U. (1985) Amer. J. Puthol. 121,382-391. 4. Fukuda, M. (1985) Biochim. Biophys. Actu 780, 119-150. 5. Feizi, T., and Childs, R. A. (1985) Trends Biol. Sci. 10, 6. Malaise, Mahieu,

M. G., Franchimont, P. R. (1987) Clin.

T. W. Annu.

in Protein Humana,

A., and Frederick,

B. (1988)

27. Maley, F., Trimble, R. B., Tarentino, A. L., and Plummer, Jr. (1989) Anal. B&hem. 180, 195-204. 28. Dell, A. (1987) Adu. Carbohydr. Chem. Biochem. 45,19-72. 29. Edge, A. S. B., Van Langenhove, (1986) Biochemistry 25,8017-8024. 30. Angel,

A.-S.,

Lindh,

F., and

A., Reinhold,

Nilsson,

B. (1987)

SeClif-

T. H.,

V., and Weber, Carbohydr.

P. Res.

168.15-31. 24-29.

P., Gomez, F., Bouillenne, C., and Zmmurwl. Immunopathol. 46,1-16.

31. Sasaki,

H., Ochi,

N., Dell, A., and Fukuda,

M. (1988)

Biochemistry

27,8618-8626. 32. Hernandez,

L. M.,

Ballou,

L., Alvarado,

E., Gillece-Castro,

B. L.,

BARR ET AL. Burlingame, 11,849-11,856.

33. Poulter, (1989)

A. L., and

Ballou,

C. E. (1989)

L., Earnest, J. P., Stroud, Proc. Natl. Acad. Sci. USA

34. Parekh,

R. B., Tse, A. G. D., Dwek, Rademacher, T. W. (1987) EMBOJ

J. Biol.

Chem.

264,

41. Brauman, J. Amer.

A. L.

42. Eagles, J., Laird, Mass Spectrom.

R. M., and Burlingame, 86,6645-6649. R. A., Williams, 6, 1233-1244.

A. F., and

35. Sharma, S. K. (1990) Adu. Drug Delivery Rev. 4, 87-111. 36. Ciucanu, I., and Kerek, F. (1984) Carbohydr. Res. 131, 209-217. 37. Hardy, M. R., and Townsend, R. R. (1988) Proc. Natl. Acad. Sci. USA

a&3289-3293.

38. Lee, Y. C. (1990) Anal. Biochem. 189, 151-162. 39. Wang, W. T., and Zopf, D. (1989) Carbohydr. Res. 189, 40. Kobata, A. (1979) An&. Biochem. 100, 1-14.

l-11.

J. I., Bryson, J. A., Kahl, Chem. Sot. 92,6679-6680.

W. M., Self, R., and Synge, T., and Rauvala,

44. Hakomori,

S. (1964)

45. Spellman,

M. W. (1990)

C.-T.,

N. J. (1970)

L. M. (1974)

Biomed.

Carbohydr.

Res. 80,

1, 43-48.

43. Finne, J., Krusius, 336-339.

46. Yuen,

D. C., and Nelson,

Carr,

J. Biochem. Anal.

H. (1980) (Tokyo)

Chem.

S. A., and Feizi,

62,

55,

205-208.

1714-1722.

T. (1990)

Eur.

J. Biochem.,

192.523-528. 47. Covey, T. R., Bonner, R. F., Shushan, B. I., and Henion, J. (1988) Rapid Commun. Mass Spectrom. 2, 249-255. 48. Hemling, M. E., Roberts, G. D., Johnson, W., and Carr, S. A. Biomed. Environ. Mass Spectrom., in press.