a multi-method approach toward de novo glycan characterization: a

Glycobiology vol. 20 no. 5 pp. 629–647, 2010doi:10.1093/glycob/cwq012Advance Access publication on January 27, 2010

A multi-method approach toward de novo glycancharacterization: A Man-5 case study

Justin M. Prien1, Bradley D. Prater2,and Steven L. Cockrill2

2Analytical Sciences, Amgen, Inc., 4000 Nelson Rd., Longmont, CO 80503,USA

Received on November 15, 2009; revised on January 20, 2010; accepted onJanuary 21, 2010

Regulatory agencies’ expectations for biotherapeutic ap-proval are becoming more stringent with regard toproduct characterization, where minor species as low as0.1% of a given profile are typically identified. The missionof this manuscript is to demonstrate a multi-method ap-proach toward de novo glycan characterization andquantitation, including minor species at or approachingthe 0.1% benchmark. Recently, unexpected isomers of theMan5GlcNAc2 (M5) were reported (Prien JM, Ashline DJ,Lapadula AJ, Zhang H, Reinhold VN. 2009. The highmannose glycans from bovine ribonuclease B isomer char-acterization by ion trap mass spectrometry (MS). J Am SocMass Spectrom. 20:539–556). In the current study, quanti-tative analysis of these isomers found in commercial M5

standard demonstrated that they are in low abundance(<1% of the total) and therefore an exemplary “litmus test”for minor species characterization. A simple workflow de-vised around three core well-established analyticalprocedures: (1) fluorescence derivatization; (2) online rap-id resolution reversed-phase separation coupled withnegative-mode sequential mass spectrometry (RRRP-(−)-MSn); and (3) permethylation derivatization with nanos-pray sequential mass spectrometry (NSI-MSn) providescomprehensive glycan structural determination. All meth-ods have limitations; however, a multi-method workflow isan at-line stopgap/solution which mitigates each method’sindividual shortcoming(s) providing greater opportunityfor more comprehensive characterization. This manu-scr ipt i s the f i rs t to demonstrate quant i ta t ivechromatographic separation of the M5 isomers and theuse of a commercially available stable isotope variant of2-aminobenzoic acid to detect and chromatographicallyresolve multiple M5 isomers in bovine ribonuclease B.With this multi-method approach, we have the capabili-ties to comprehensively characterize a biotherapeutic’sglycan array in a de novo manner, including structuralisomers at ≥0.1% of the total chromatographic peakarea.

Keywords: glycan/glycoprotein /mass spectrometry /monoclonal antibody/oligosaccharide

Introduction

Significant advancements in methodologies for glycan structur-al analysis have been achieved in recent years, particularlywith respect to chromatography and mass spectrometry (Ash-line et al. 2005; Harvey 2005a, b, c; Anumula 2006; Alvarez-Manilla et al. 2007; Costello et al. 2007; Pabst et al. 2007; Prat-er et al. 2009). Even with such advancement, the complexbranching patterns and intrinsic isomeric characteristics of gly-can structures still pose significant analytical challenges,particularly with respect to the quantitation and characteriza-tion of low abundant species.During the development and commercialization of biothera-

peutics, the detection, characterization and justification ofminor species is necessary regardless of origin. The detailedcharacterization of a biotherapeutic’s glycan array is a regula-tory requirement (ICH 2006; EMEA 2007); however, no lessimportant is the characterization of glycan artifacts or modifi-cations spurred during drug processing and/or productcharacterization (e.g., buffer exchange, enzymatic release, pu-rification, etc.). Consequently, de novo characterizationcapabilities are needed. Significant effort is made to fully char-acterize the minor peaks in a chromatographic profile fordeeper understanding of the product and the potential impactthat a change to product development procedure, particularlychanges in manufacturing site or scale, might have on thatproduct. As such, repeated de novo characterization is likelyduring the lifecycle of a glycosylated biotherapeutic.Traditional elucidation of glycan structures has involved iso-

lation of individual glycan species by chromatographicseparation followed by off-line matrix-assisted laser desorptionionization time-of-flight analysis for basic mass determinationand serial exoglycosidase digestions for glycan characteriza-tion (Lee 1990; Sutton et al. 1994; Anumula et al. 1998).Many drawbacks are associated with enzymatic sequencingsuch as incomplete digestion, variation in enzyme activity,and enzyme purity, specificity and availability, all of whichgreatly confound analysis and can lead to misinterpretationand/or incomplete understanding of a biotherapeutic’s chro-matographic profile (Sutton et al. 1994; Harvey et al. 2008).Separation methods historically have included high pH anionexchange chromatography (HPAEC) (Lee 1990) or normal-phase high performance liquid chromatography (NP-HPLC)(Guile et al. 1996). More recently, separations based upon al-ternative modes, including reverse-phase (RP-HPLC) and

© The Author 2010. Published by Oxford University Press. All rights reserved. For permissions, please e-mail: [email protected] 629

1To whom correspondence should be addressed: Fax: +(303) 401-4404; e-mail:[email protected]

Downloaded from https://academic.oup.com/glycob/article-abstract/20/5/629/1987998by gueston 10 April 2018

porous graphitic carbon have been implemented (Kopp.K1997; Delaney et al. 2001; Kawasaki et al. 2003; Ninonuevoet al. 2005; Chen and Flynn 2007; Costello et al. 2007; Pabstet al. 2007; Chen et al. 2009; Prater et al. 2009).

In recent years, mass spectrometry-based techniques for gly-can structural elucidation have been developed (Reinhold et al.1995; Viseux et al. 1997; Weiskopf et al. 1997; Brull et al. 1998;Reinhold and Sheeley 1998; Viseux et al. 1998; Weiskopf et al.1998; Harvey et al. 2002; Pfenninger et al. 2002a; Ashline et al.2005; Harvey 2005b; Geyer and Geyer 2006). These methodsare analogous to the more established approaches in the proteo-mics arena, provide exquisite sensitivity, rapid data acquisitionand fundamental physical measurement-based analysis andhave become an attractive alternative to traditional exoglycosi-dase sequencing.

A logical extension of glycan analysis is to couple themass spectrometric analysis with the separation itself. Onlineliquid chromatography-mass spectrometry (LC-MS) can be apowerful analytical tool for rapid quantitation and identifica-tion of glycan species. As noted earlier, HPAEC has longbeen used to separate glycans into various groups based oncharge. With appropriate online desalting, HPAEC can be in-terfaced with a mass spectrometer creating a method quiteadvantageous for glycoproteins with highly charged glycanspecies (Thayer et al. 1998; Bruggink et al. 2005; Chataigneet al. 2008). An alternative separation mode, NP-HPLC, usesamide-based columns to separate glycans based on size andhydrophilicity (Guile et al. 1996; Wuhrer et al. 2004) and hasalso been successfully coupled to mass spectrometry (MS)detection (Wuhrer et al. 2004). Porous graphitized carbonchromatography is quite compatible with mass spectrometricconditions and demonstrates utility for resolving glycans ofsimilar structure (Kawasaki et al. 2003; Ninonuevo et al.2005; Costello et al. 2007; Pabst et al. 2007). Reverse-phasechromatographic methods demonstrate the capability for res-olution of structural isomers as well as segregation by glycanclass (Kopp and Werner 1997; Delaney and Vouros 2001;Chen and Flynn 2007; Chen et al. 2009; Prater et al. 2009)and are intrinsically compatible with MS detectors. Recent lit-erature reports have illustrated the implementation of moreelaborate MS methodologies such as online MSn schemes(Delaney and Vouros 2001; Chen and Flynn 2007; Chen etal. 2009).

In general, upfront separation improves ionization efficien-cy and helps reduce the complexity of mass spectra;unfortunately, the potential for coelution of discrete glycanspecies afflicts all chromatographic methods. Coeluting glyco-mers can be distinguished according to differences in theirintact mass; however, the presence of coeluting positional iso-mers requires additional analysis via subsequent rounds ofsequential disassembly.

No single current analytical methodology can independentlyacquire all the qualitative and quantitative data necessary tocompletely characterize a glycan topology. For example, posi-tive-mode analysis of reducing-end derivatized protonatedglycans is readily compatible with upfront reverse-phase sepa-ration. Unfortunately, data quality is compromised bymonomer rearrangement and the typical absence of cross-ringfragmentation, limiting structural elucidation to monosaccha-ride sequencing (Brull et al. 1998; Franz and Lebrilla 2002;

Harvey et al. 2002). Informative cross-ring fragmentation doesoccur in this mode using metal-adducted ions; however, thepresence of “internal residue” fragments stemming from mul-tiple independent cleavages promotes isobaric as well asisomeric product ions making structural determination nebu-lous (Brull et al. 1998; Franz and Lebrilla 2002; Harvey2005a, 2008; Wuhrer et al. 2006).

Negative-mode analysis of reducing-end derivatized glycansprovides more informative structural data than correspondingpositive-mode native analysis (Chai et al. 2001, 2002, 2006;Pfenninger et al. 2002a, b; Harvey 2005b, c, d; Harvey et al.2008). Unexpectedly, we report that under rapid resolutionreversed-phase chromatography (RRRP) low-energy collision-induced dissociation (CID) negative-mode conditions, “internalresidue” fragment formation is observed similar to positive-mode analysis. This greatly complicates spectral interpretation,rendering structural isomer differentiation and de novo charac-terization difficult.

Orthogonal analysis via permethylation and nanospray se-quential mass spectrometry (NSI-MSn) provides the mostinformative spectral data for structural elucidation (Ashline etal. 2005, 2007; Alvarez-Manilla et al. 2007; Costello et al.2007; Prien et al. 2008, 2009). However, permethylated anal-ysis requires an additional derivatization step and is not well-suited for certain glycan conjugates, such as sulfated glycosa-minoglycans (Zaia et al. 2007).

A simple workflow devised around three well-establishedcore analytical procedures: (1) fluorescence derivatization;(2) online separation coupled with negative-mode sequentialmass spectrometry; and (3) permethylation derivatization withNSI-MSn provides a more comprehensive approach toward thecharacterization of a glycoprotein’s glycan array, including mi-nor structural isomers. The use of permethyl derivatization andNSI-MSn is the critical aspect for glycan structural character-ization. While the aforementioned mass spectrometric methodshave been previously described as stand-alone techniques (Do-mon and Costello 1988; Reinhold et al. 1995; Weiskopf et al.1997; Reinhold and Sheeley 1998; Sheeley and Reinhold1998; Weiskopf et al. 1998; Pfenninger et al. 2002a, b; Ciuca-nu and Costello 2003; Ashline et al. 2005, 2007; Harvey2005a, b, c, d, e; Kang et al. 2005, 2008; Lapadula et al.2005; Anumula 2006; Chen and Flynn 2007; Costello et al.2007; Harvey et al. 2008; Prien et al. 2008, 2009; Chen etal. 2009), an inclusive workflow coupling these methodologieshas not been described, and its demonstration will be of broadutility for the biotherapeutic industry.

Results

The oligosaccharide Man5GlcNAc2 (M5) has been analyzed indetail by a multitude of analytical methodologies providing aforum to draw comparison between the individual analyticalmethods and perform benefit/shortcoming analysis (Fu et al.1994; Harvey 2005a, c, d; Costello et al. 2007; Zhuang et al.2007; Zhao et al. 2008; Prien et al. 2009). Here, we demon-strate both quantitation and structural characterization ofthree M5 structural topologies present in commercially avail-able M5 standard derived from porcine thyroglobulin as wellas chromatographic separation of multiple M5 isomers in bo-vine RNase B.

630

JM Prien et al.


NomenclatureThe nomenclature, used herein to identify the intact glycantopologies, is based on the number system for high mannose(Man) isomers as described previously (Prien et al. 2009).Briefly, three distinct M5 topologies were annotated T1, T2

and T3. Additional linear notation is utilized for product ionspectral annotation. We apologize for the introduction of addi-tional notation beyond the established (Domon and Costello1988). Our goal for this nomenclature, however, is to afford an-notation of product ions without assigning an inferred structure.

631

Fig. 1. Fluorescence profiles and single stage mass spectra from online RRRP-(−)-mode MS analysis of 2-AA and 2-AB derivatized oligomannose-5 (M5) standardpurified from porcine thyroglobulin. Assigned structures in (A) and (F) were confirmed by (−)-mode MSn and permethylated NSI-MSn as detailed in later sections.

Comprehensive analysis of Man-5 positional isomers


This will help reduce misinterpretation due to the enormous iso-meric and isobaric fragment ion overlap intrinsic to MS glycananalysis. A complete description of nomenclature used is dis-cussed in the Materials and methods section.

Online RRRP-(−)-mode MS of M5 standard from porcinethyroglobulinFour peaks with mass values consistent with that expected forfluorescently labeled M5 (m/z 1354.7 and m/z 1353.4 for 2-

aminobenzoic acid (2-AA) and 2-aminobenzamide (2-AB) la-bels, respectively) were resolved and quantitated duringsample purity assessment using RRRP-(−)-mode MS(Figure 1A–J). In addition, a single M6 structure was observed,presumably present due to co-purification during commercialpreparation. As evidenced in Table I, the putative M5 structurecomposes ~96% of the total peak area representing the M5 pool(Figure 1A and F, peak (4)), while the three additional M5 iso-mers collectively represent ~4.0% of the total M5 peak area(Figure 1A and F, peaks (1), (3) and (5)).

632

Table I. Comparison of RRRP-HPLC methods for the quantitative analysis of 2-AA and 2-AB derivatized oligomannose-5 (M5) standard from porcine thyro-globulin. Error is expressed as standard deviation from 15 independent analyses. Assigned structures were confirmed by RRRP-(−)-mode MSn and permethylatedNSI-MSn. The “T” notation corresponds to the different topologies (as indicated in the structural depictions), elucidated in the final results.

Sample % of total M5 pool

Peak (1)

T1*

Peak (3)

T2

Peak (4)

T1

Peak (5)

T3

Porcine

Thyroglobulin

M5-2AA 3.3 ± 0.2 0.7 ± 0.1 95.9 ± 0.2 0.1 ± 0.0

M5-2AB 2.9 ± 0.1 0.6 ± 0.1 96.4 ± 0.2 0.1 ± 0.0

Fig. 2. Static nanospray negative-mode MS2 spectra of the [M-H]− ion from (A) 2-AA derivatized, m/z 1354.7, and (B) 2-AB derivatized, m/z 1353.7,oligomannose-5 (M5) standard purified from porcine thyroglobulin.

JM Prien et al.


In a previous report, we identified the presence of an earlyeluting M5 isomer, which is consistent with the peak (1) reten-tion time relative to peak (4) (Prater et al. 2009). Recently,epimerization of the core N-acetylglucosamine (GlcNAc) resi-due during PNGase F release was reported as a possible originof certain structural isomers (Liu et al. 2009), and our analysisis indicative that the structural differences between peaks (1)and (4) are restricted to linkage or residue identities withinthe putative oligosaccharide core. A detailed investigation ofthis structure has recently been completed (to be published).This report, therefore, is focused on the structural elucidationof the remaining three resolved M5 peaks denoted as peaks (3),(4) and (5) (Figure 1).

Released glycans are exposed to harsh reaction conditionsduring fluorescence derivatization, e.g., heating at 65°C inthe presence of acetic acid for 2 h. These conditions might leadto decomposition or structural rearrangement of a glycan’s na-tive structure producing low abundant isomers. To address thisconcern, we performed a study exposing M5 to the harsh label-ing conditions over a time course (2–18 h). If the derivatization

conditions lead to isomeric formation, one would expect thepeak area for each isomeric species to elevate over time. Thestudy revealed that no changes were observed in the relativepeak areas of the T1, T2 and T3 isomeric species, indicatingthat the labeling conditions do not lead to structural rearrange-ment (data not shown).

Impact of fluorescence derivatization on separation and massspectral qualityReducing-end fluorescent derivatization is a widely used meth-odology (Anumula 2006; Siemiatkoski et al. 2006). Certainfactors must be considered when choosing the appropriate re-ducing-end label, including label reactivity and compatibilitywith the desired chromatographic and mass spectrometricmethods.From an examination of the fluorescent profiles (Figure 1A

and F), the most significant difference between the 2-AA and 2-AB labels relates to the shorter separation time provided by the2-AB label using identical separation conditions. All five peaks

633

Fig. 3. Online RRRP-(−)-mode MS/MS spectra of the [M-H]− ion from 2-AA derivatized M5 (M5-2-AA), m/z 1354.7, extracted from various resolvedchromatographic peaks (3), (4) and (5). Product ions are indicated on in each spectrum and their proposed origins depicted on the associated cartoon.



are observed in both profiles and baseline resolved. However,the advantageous speed of separation provided by the 2-AB la-bel is compromised by the quality of mass spectral data; anobvious difference between the intact mass profiles of 2-AAand 2-AB labeled M5 is immediately evident (Figure 1B–Eand G–J). M5-2-AB species tend to ionize as singly chargedspecies and prefer the adduction of acetate (+59 Da). In partic-ular, the adduction of two acetate ions (+118 Da), m/z 1467.5[M-3H + 2CH3COOH]

− (Figure 1G–J). Conversely, M5-2-AAis primarily preserved as deprotonated species [M-H]−, m/z1354.7 (Figure 1B–E), resulting in improved spectral qualityfor the 2-AA labeled substrate.

For our characterization purposes, the 2-AA label confersmany advantages over the more commonly used 2-AB label,such as reactivity in aqueous derivatization conditions improv-ing sample workflow efficiency, and an increase in fluorescenceintensity improving chromatographic sensitivity for low abun-dant glycan species (Anumula and Dhume 1998; Anumula2006). Intriguingly, 2-AA affords even greater advantagewith respect to mass spectrometric analysis. Under negative-mode conditions, preferential deprotonation occurs at the car-boxyl group of the 2-AA label producing predictable Y-ionfragmentation (Harvey 2005a, c). As such, this fixed deprotona-tion promotes efficient MSn disassembly from the nonreducingend to the reducing end of the glycomer allowing sequencingwithout the burden of structural rearrangement as observedin positive-mode analysis (Brull et al. 1998; Franz and Leb-rilla 2002; Chen and Flynn 2007; Harvey et al. 2008). In ourexperience, 2-AB labeled glycan tend to “partially permethy-late” (data not shown). In contrast, the 2-AA label appearsto readily permethylate and demonstrates a single intactmass species. Having a single intact mass reduces speciesdilution and spectral profile complexity resulting in im-proved ion intensity, sensitivity and spectral quality, whichis advantageous for de novo characterization of low abun-dant species. Additionally, 2-AA is available as a 13[C6]stable isotope analog.

Direct infusion of fluorescent labeled M5 standard fromporcine thyroglobulinDirect infusion of the unseparated 2-AA and 2-AB labeled M5

standard was executed in negative mode to assess whether theapparent M5 positional isomers (peaks (1), (3), (4) and (5) inFigure 1A and F) were detectable without separation or furthersample processing.

As shown in Figure 2A, 2-AA labeled M5 produces Y-ionfragments, m/z 1192.4 (−162 Da; H4N1n-2AA-[OH]; Y4α′,Y4α″, or Y3β) and m/z 868.4 (−486 Da; H2N1n-2AA-[OH];Y3α or Y3β/Y4α′/Y4α″), which are consistent with the neutralloss of one and three mannose residues, respectively, most likelycorresponding to the loss of the 3-antenna and 6-antenna of the

canonicalM5 topology (annotated as T1). In addition, diagnosticions corresponding to the 6-antenna are observed: m/z 647.3(H4-[eneOH]; “D-ion” or C3/Z3β) and m/z 503.3 (H3-[OH];C2α-ion) (Harvey 2005b, c).

Interestingly, a low abundant ion, m/z 1030.4 (−341 Da;H3N1n-2AA-[OH]; Y3β/Y4α′; Y3β/Y4α″; or Y4α′/Y4α″), is pres-ent in the MS2 spectrum of 2-AA labeled M5 (Figure 2A). Thepresence of this ion is inconsistent with the predicted negative-mode fragmentation of the T1 M5 structure and has been pre-viously interpreted as an “internal residue” product ion fromthe neutral loss of two independent mannose residues (Harvey2005a); however, this ion may also be a Y-ion fragment fromadditional positional isomers present in the M5 standard. Add-ing further to the complexity, a combination of both of theabove hypotheses forms an equally valid third explanation.Product ion m/z 485.2 is another low abundant fragment thatmay be generated from multiple origins: (1) the loss of water(−18) from the C2α ion, m/z 503.2, of the putative M5 structure;(2) a D-ion fragment from another isomer that is present; or (3)an isomeric mixture of both (1) and (2) (Supplemental data,Figure S1). Similarly, most product ions in the 2-AB labeledM5 spectrum (Figure 2B) are consistent with the canonicalM5 topology, but again two product ions, m/z 485.2 and m/z1029.3, are inconsistent with the expected. Hence, without sep-aration of the positional isomers and the ability to differentiatenative hydroxyls from those formed during fragmentation, de-finitive assessment of ion origin cannot be made.

An unexpected observation was that under low energy condi-tions a series of sequential neutral hexose loss (−162 Da)beginning with a 2,4A cross-ring fragment is observed: m/z869.3 (H5-[

2,4A]) → m/z 707.3 (H4-[2,4A]) → m/z 545.2 (H3-

[2,4A])→ m/z 383.2 (H2-[2,4A]) (Figure 2A and B). This neutral

loss series was previously observed only at higher fragmentationenergy levels (Harvey 2005c). This series exemplifies sequential“internal residue” formation stemming from multiple indepen-dent cleavages. As a result, it is plausible that the observedproduct ions m/z 545.2 andm/z 383.2 may be derived from mul-tiple internal cleavages from different regions of the molecule.For example, m/z 545.2 may represent an outer arm 0,4A2α-ion(or H3-[

0,4A]) or the “internal residue”H3-[2,4A] ion described in

the above neutral loss series. If multiple M5 structures are pres-ent, the number of isomeric fragments m/z 545.2 representsbecomes even more confounding.

Given the information gleaned from the chromatographicseparation indicating the presence of multiple M5 topologies(Figure 1A and F) as well as the presence of potentially incon-sistent ions (m/z 485.2, 1029.3, and 1030.4) and “internalresidue” product ion formation in the unseparated (−)-modeMS/MS spectra (Figures 2A and B), an assessment of localiza-tion of these inconsistent ions to a particular chromatographicpeak was undertaken.

634

Fig. 4. NSI-MSn spectra for peak (4) M5-2-AA fraction. Sequential disassembly of the putative M5 structure, T1, prepared as a permethylated 2-AA derivatizedsodium adducted species. The MSn pathway followed appears as inserts in the top left corner of the spectra (A–E). Only one of many isomeric partial structures ispresented as an associated cartoon. The cartoons are simply used to demonstrate product ion origin specific to the cartoon shown and a single MSn fragmentationpathway unique to the topology shown in (A). Multiple cartoons and MSn fragmentation pathways specific to the intact topology could be shown for each spectrum.To the far right of each spectrum, selected product ions are annotated. Please be advised that not all selected product ions correspond to the associate cartoon;however, the product ions do correspond to the compilation of isomeric precursor ions (e.g., partial structures) that are simultaneously isolated and disassembled.Importantly, all product ions annotated and cartooned originate from the intact species cartooned in (A). Product ions highlighted in bold print were selected forsubsequent dissociation and constitute the MSn pathway.

JM Prien et al.


635Figure 4.



636

Table II. Permethylated MSn pathways and topologies for 2-AA derivatized sodium adducted oligomannose-5 Standard (M5) from porcine thyroglobulin. The tablealso includes a key to the various fragment symbols used in the table. Spectral annotation and topology assignments were performed by gtSuite™

M5-2-AA

Putative (T1) T2 T3

MSn Pathways Consistent Topologies

a. 1728.7 → 1302.6 (loss of ) → 1084.4 (loss of ) T1, T2, T3

b. 1728.7 → 1302.6 (loss of ) → 1084.4 (loss of )→ 866.4 (loss of )

T1, T2, T3

c. 1728.7 → 1302.6 (loss of ) → 1084.4 (loss of )→ 866.4 (loss of ) 648.2 (loss of )

T1

d. 1728.7 → 1302.6 (loss of ) → 1084.4 (loss of ) → 866.4 (loss of ) 648.2 (loss of ) → 458.1 ( )

T1

e. 1728.7 → 1302.6 (loss of ) → 1084.4 (loss of ) → 866.4 (loss of ) 648.2 (loss of ) → 421.3

T1

f. 1728.7 → 1302.6 (loss of ) → 1084.4 (loss of ) → 866.4 (loss of ) 648.2 (loss of ) → 403.2

T1

g. 1728.7 → 1302.6 (loss of ) → 1084.4 (loss of ) → 866.4 (loss of ) → 648.2 (loss of ) → 268.1

T1

h. 1728.7 → 1510.7 (loss of ) T1, T2, T3

i. 1728.7 → 1510.7 (loss of ) → 1292.6 (loss of ) T1, T2, T3

→

→

→

→

j. 1728.7 → 1510.7 (loss of ) → 1292.6 (loss of ) → 1074.5 (loss of )

T1

k. 1728.7 → 1510.7 (loss of ) → 1292.6 (loss of ) → 1074.5 (loss of ) → 648.2 (loss of )

T1

l. 1728.7 → 1302.6 (loss of ) → 1084.4 (loss of ) → 667.4

( , C3-type ion)

T1, T3

m. 1728.7 → 1302.6 (loss of ) → 1084.4 (loss of ) → 737.3 3,5A4- ion)

T1, T3

n. 1728.7 → 1302.6 (loss of ) → 1084.4 (loss of ) → 880.4 (loss of )

T2, T3

o. 1728.7 → 1302.6 (loss of ) → 1084.4 (loss of ) → 880.4 (loss of )→ 662.1 (loss of )

T2, T3

(

JM Prien et al.


Online RRRP-(−)-mode MS/MS analysis of M5 standard fromporcine thyroglobulinBoth 2-AA and 2-AB derivatized M5 standards were used dur-ing comprehensive analysis; however, only the 2-AA labeledM5 species, because of the advantages mentioned above, willbe subsequently discussed.

Online RRRP-(−)-mode MS/MS: peak (4), ~96% abundanceOnline LC-MS analysis of peak (4) revealed a parent ion of m/z1354.7 for the 2-AA labeled M5 (Figure 3A). The observation

of diagnostic ions m/z 647.3, m/z 629.3, m/z 575.3 and m/z503.3 confirm the presence of three mannose residues on the6-antenna and is indicative of the T1 branching pattern(Figure 3A). Interestingly, the inconsistent product ion m/z1030.4 and the sequential neutral loss series beginning withthe 2,4A cross-ring fragment ion m/z 869.2 are observed inthe peak (4) MS/MS spectrum. Equally important is the ab-sence of the m/z 485.2 product ion. These findings suggestthat m/z 1030.4 represents the loss of two independent man-nose residues consistent with previous hypotheses (Harvey2005a), while the sequential neutral loss series indicates that

637

Table II (continued).

p. 1728.7 → 1302.6 (loss of ) → 1084.4 (loss of ) → 880.4 (loss of )→ 662.1 (loss of ) → 458.1 ( )

T2, T3

q. 1728.7 → 1302.6 (loss of ) → 1084.4 (loss of ) → 880.4 (loss of )→ 662.1 (loss of ) → 458.1 ( ) → 268.1

T2, T3

r. 1728.7 → 1302.6 (loss of ) → 1084.4 (loss of ) → 880.4 (loss of ) → 463.2( ,C2-type ion)

T2

s. 1728.7 → 1302.6 (loss of ) → 1084.4 (loss of ) → 880.4 (loss of ) → 445.2 ( , B2-type ion)

T2

t. 1728.7 → 1302.6 (loss of ) → 1084.4 (loss of ) → 880.4 (loss of ) → 676.3 (loss of )

T3

u. 1728.7 → 1302.6 (loss of ) → 1084.4 (loss of ) → 880.4 (loss of ) → 676.3 (loss of ) → 458.1 ( )

T3

v. 1728.7 1302.6 (loss of ) → 1084.4 (loss of ) 880.4 (loss of ) → 676.3 (loss of ) → 241.1 (loss of )

T3

w. 1728.7 → 1302.6 (loss of ) → 1084.4 (loss of ) → 737.3 3,5A4- ion) → 667.3 ( , C3-type ion) → 445.2 ( , B2-type ion)

T3

v. 1728.7 → 1302.6 (loss of ) → 1084.4 (loss of ) → 737.3 ( 3,5A4-

ion) → 667.3 ( , C3-type ion) → 463.2 ( , C2-type ion)

T3

→→

(

Symbol Fragment identity

Neutral loss(singly-

chargedprecursor)

Sodiated mass [M+Na]+

Terminal Mannose, B type 218 241

Internal Mannose, B/Y type 204 227

Internal Mannose, C/Y type 222 245

Branched Mannose, B/Y/Y type 190 213

Terminal B2-type ion 422 445

Terminal C 2-type ion 440 463

Internal C/Y type ion 426 449

Terminal C 3-type ion 644 667

Core Man- β1,4-GlcNAc, B/Y/Y type

notapplicable

458

Internal HexNAc, B/Y type 245 268

Reducing-end GlcNAc-2AA, Y-type

426 not usuallyseen



“internal residue” formation occurs similar to the positivemode.

Online RRRP-(−)-mode MS/MS: peak (3), ~0.7% abundanceThe “D-ion” is a diagnostic of gross structural features such as6-antennal branching (Harvey 2005c). Specifically, the forma-tion of this ion originates from neutral loss of the 3-antennaand loss of the penultimate and reducing-end GlcNAc-GlcNAcdi-N-Acetyl-D-glucoaminosyl moiety. The “D-ion” generatesprevalent product ions corresponding to the loss of water (D′)and subsequent 0,3A-ion and C-ion, (C/Z −18 Da), (C/Z −72Da) and (C/Z −144 Da), respectively. These diagnostic ionsare observed in the MS/MS spectrum for peak (3) (Figure 3B),albeit with one less mannose residue than observed for the T1topology of peak (4) (Figure 3A). As shown in Figure 3B, prod-uct ions m/z 485.3 (H3-[eneOH]; “D-ion” or C3/Z3α), m/z 467.3(H3-[ene2OH]; “D′-ion” or C3/Z3α-18 Da) and m/z 413.3 (H2-[0,3A]; 0,3A2-ion) correspond to two mannose residues presenton the 6-antenna suggesting gross antennary structural differ-ences between the M5 species present in peaks (3) and (4).Moreover, ions representative of the three mannose residuesconcomitant with the T1 6-antennae (i.e., m/z 647.3, 629.3,and 575.3) are notably absent (Figure 3B). As such, the presenceand increased relative abundance ofm/z 1030.4 may be rational-ized in an entirely different manner. Here, the product ion isunlikely to be an “internal residue” fragment ion but the production of a single cleavage event occurring between a terminaldi-mannosyl group (−341 Da; H3N1n-2AA-[OH]; Y3α orY3β) and the β1,4 core mannose of the proposed T2 structure(Figure 3B).

Online RRRP-(−)-mode MS/MS: peak (5), ~0.1% abundanceA low abundant peak, peak (5) in Figure 1A and E, occupying~0.1% of the total M5 fluorescent peak area presents a parention (m/z 1354.7) and MS/MS product ions that are similar, butnot identical, to the other M5 structures (Figure 3C). Notably,this species also contains the m/z 1030.4 ion. Unfortunately,the online MS/MS spectral quality is insufficient to confidentlypropose a structure (Figure 3C).

In summary, the sensitivity and resolving power of onlineRRRP-(−)-mode MS/MS and MSn (Supplemental data, FigureS2) provides conclusive evidence confirming the presence ofmultiple M5 structures. Unfortunately, the potential formationof “internal residue” fragments can lead to data ambiguity andconfound structural characterization without having a prioriknowledge of the structures that are present. That is, whilethe M5 isomers may be separated chromatographically, onlinenegative-mode analysis cannot unambiguously differentiateand characterize the structural differences of the M5 speciesin a de novo manner, thereby requiring alternative analytictools for structural annotation. Fortunately, as detailed in the

subsequent sections, infusion-based MSn of permethylated gly-cans provides the specific information necessary to assignstructural motifs.

In order to reconcile the specific structural identities of eachpeak in the fluorescent profile, peaks (3), (4) and (5) were iso-lated from the chromatographic separation, subjected topermethylation and subsequently characterized by NSI-MSn

analysis.

Permethylated NSI-MSn analysisPermethylated NSI-MSn has been discussed in detail in the lit-erature (Reinhold et al. 1995; Viseux et al. 1997; Weiskopf et al.1997, 1998; Reinhold and Sheeley 1998; Sheeley and Reinhold1998; Viseux et al. 1998; Karlsson, Schulz, et al. 2004; Karls-son, Wilson, et al. 2004; Ashline et al. 2005, 2007; Lapadulaet al. 2005; Hanneman et al. 2006; Costello et al. 2007; Prienet al. 2008, 2009). In short, several critical benefits of glycanpermethylation with respect to structural characterization andmass spectrometric analysis include: (1) increased oligosac-charide ionization efficiency (Harvey 1999); (2) increasedabundance of diagnostic cross-ring fragments which can beutilized to determine linkage position (Mechref et al. 2003,Morelle et al. 2004; Ashline et al. 2005); and (3) upon fragmen-tation, terminal and internal residues exhibit mass differences(14 Da) depending on the number of exposed hydroxyl groups,commonly known as “scars” (Ashline et al. 2007; Costello et al.2007). These attributes allow for the de novo characterizationof individual structures from unseparated glycan pools (Rein-hold et al. 1995; Viseux et al. 1997, 1998; Weiskopf et al.1997, 1998; Reinhold and Sheeley 1998; Sheeley and Rein-hold 1998; Karlsson, Schulz, et al. 2004; Karlsson, Wilson, etal. 2004; Ashline et al. 2005, 2007; Lapadula et al. 2005;Hanneman et al. 2006; Costello et al. 2007; Prien et al.2008, 2009).

Permethylated NSI-MSn analysis of M5 positional isomersStructural rearrangement and isomeric formation consequent ofchemical modification and gas-phase analysis is a persistentconcern surrounding permethyl derivatization and ion trap dis-assembly. Here, these hypotheses are nullified knowing thediscrete M5 isomers are separated by chromatography priorto permethyl derivatization and mass spectrometric conditions.

Permethylated NSI-MSn: peak (4)The signature MSn pathway, m/z 1728.7 (H5N1n-2AA) → m/z1302.6 (−426 Da; loss of 2-AA reducing-end GlcNAc; H5N1-[ene1]) → m/z 1084.4 (−218 Da; loss of terminal B1 hexose;H4N1-[ene1OH1]) → m/z 866.4 (−218 Da; loss of second ter-minal B1 hexose; H3N1-[ene1OH2]) → m/z 648.2 (−218 Da;loss of third terminal B1 hexose; H2N1-[ene1OH3]) → m/z458.2 (−190 Da; loss of B/Y/Y hexose; core Man1,4GlcNAc

638

Fig. 5. NSI-MSn spectra for peak (3) M5-2-AA collected fraction. Sequential disassembly of the T2 M5 positional isomer, prepared as a permethylated 2-AAderivatized sodium adducted species. The MSn pathway followed appears as inserts in the top left corner of the spectra (A–E). Only one of many isomeric partialstructures is presented as an associated cartoon. The cartoons are simply used to demonstrate product ion origin specific to the cartoon shown and a single MSn

fragmentation pathway unique to the topology shown in (A). Multiple cartoons and MSn fragmentation pathways specific to the intact topology could be shown foreach spectrum. To the far right of each spectrum, selected product ions are annotated. Please be advised that not all selected product ions correspond to the associatecartoon; however, the product ions do correspond to the compilation of isomeric precursor ions (e.g., partial structures) that are simultaneously isolated anddisassembled. Importantly, all product ions annotated and cartooned originate from the intact species cartooned in (A). Product ions highlighted in bold print wereselected for subsequent dissociation and constitute the MSn pathway.

JM Prien et al.


639Figure 5.



or H1N1-[ene1OH2]), used to distinguish and characterize theputative M5 structure from an unseparated glycan pool (Prienet al. 2009), was identified and followed in the separated peak(4) fraction confirming this peak as the putative T1 M5 struc-ture (Figure 4 and Table II, pathway d) (Supplemental data,Figure S3). Importantly, by chromatographically isolating theM5 isomers prior to permethylated NSI-MSn characterization,the ions inconsistent with the T1 structure previously observedin the unseparated pool, m/z 880.4 (H3N1-[ene1OH1]), m/z662.3 (H2N1-[ene1OH2]), m/z 463.2 (C2-ion; H2-[OH1]) andm/z 445 (B2-ion; H2-[ene1]), are notably absent from theMS2-6 spectra of isolated peak (4) (Figure 4) (Supplementaldata, Figure S3).

Permethylated NSI-MSn: peak (3)Figure 5A–E shows a MSn spectral set and MSn pathwaydistinct from T1: m/z 1728.7 (H5N1n-2AA) → m/z 1302.6(−426 Da; loss of 2-AA reducing-end GlcNAc; H5N1-[ene1]) → m/z 1084.4 (−218 Da; loss of terminal B1 hexose;H4N1-[ene1OH1]) → m/z 880.4 (−204 Da; loss of internal B/Y hexose; H3N1-[ene1OH1]) → m/z 662.3 (−218 Da; loss ofsecond terminal B1 hexose; H2N1-[ene1OH2]) → m/z 458.2(−204 Da; loss of B/Y hexose; core Man1,4GlcNAc orH1N1-[ene1OH2]). Interestingly, the ions inconsistent withthe T1 topology observed in MSn pathway for the unseparat-ed glycans, m/z 880.4, m/z 662.3, m/z 463.2 and m/z 445.2(Supplemental data, Figure S3) are abundant in the MSn spec-tral set for chromatographically separated peak (3) (Figure 5).The MS5 m/z 880.4 spectrum exhibits product ions m/z 533.3(H2-[

3,5A]), m/z 463.2 (C2-ion; H2-[OH1]) and m/z 445.2 (B2-ion; H2-[ene1]) corroborating with a fully permethylatedMan-Man disaccharide 1→6 linked to the β1,4 core mannose.The presence of a terminal B2-ion and C2-ion stemming from them/z 880.4 precursor indicates that the single exposed hydroxylgroup is located on the β1,4 Man-GlcNAc core and confirmsthat peak (3) demonstrates the T2 topology (Figure 5 and TableII, pathways r–s).

Antennal sequencing and linkage determination can be per-formed by following specific MSn fragment ion pathways thatinclude appropriate cross-ring fragments. For example, the ter-minal mannosyl residue on the 6-antenna of the T2 structure wasconfirmed as 1→3 linked through observation of MS7 productionm/z 343.3 ion following the MSn fragmentation pathway:m/z 1728.7 (H5N1n-2AA) → m/z 1302.6 (−426 Da; loss of 2-AAreducing-end GlcNAc; H5N1-[ene1]) → m/z 1084.4(−218 Da;loss of terminal B1 hexose; H4N1-[ene1OH1]) → m/z 880.4(−204 Da; loss of internal B/Y hexose; H3N1-[ene1OH1]) →m/z 533.3 (3,5A2-ion; H2-[

3,5A]) → m/z 445.2 (terminal B2-ion; H2-[ene1]) (Supplemental data, Figure S4). The MS7

spectrum of m/z 445.2 represents a fully permethylated terminalMan-Man B2-ion. The presence of a fully permethylated B1-ion(m/z 241.0) and C1-ion (m/z 259.0) is consistent with a terminal

Man-Man topology, while the presence of diagnostic cross-ringfragment ions, m/z 343.3 (3,5X-ion; H1-[

3,5Xene1]) and m/z371.4 (0,4A-ion; H1-[

0,4Xene1]) indicates that the mannose resi-dues are 1→3 linked (Ashline et al. 2005).

Permethylated NSI-MSn: peak (5)A major advantage of nanospray and direct infusion is the ex-quisite sensitivity and the ability to signal average overextended infusion times. These attributes allow for the acquisi-tion of high quality spectral data even for a low abundantspecies, such as the structure eluting at peak (5) (Figure 6).The MSn pathways followed to elucidate the peaks (3) and(5) structures are very similar, diverging only after the MS5

m/z 880.4 stage (Figures 5 and 6, Table II, pathways t and s).Although the MS5 m/z 880.4 ions from peaks (3) and (5) arecomposed of the same composition and “scar” count, H3N1-[ene1OH1], the m/z 880.4 product ion for peak (5) representsa different compilation of structures compared to the peak (3)T2 pathway (Figure 5D). In the peak (5) pathway, production m/z 676.3 of the MS5 m/z 880.4 spectra is consistent witha B/Y-ion having a H2N1 composition with a single exposedhydroxyl group “scar” (Figure 6D). Additionally, the presenceof a fully permethylated terminal C1-ion, m/z 259.1, in theMS6 m/z 676.3 spectrum indicates that the single exposed hy-droxyl group is located on the β1,4 Man-GlcNAc core,confirming peak (5) as having a unique structure, termedT3 (Figure 6E). Product ion m/z 676.3 is absent in the MS5

m/z 880.4 spectrum from the T2 structure (Figure 5D).Therefore, product ion m/z 676.3 from an MS5 m/z 880.4precursor ion is diagnostic to structure T3 (Table II, pathwayt–v).

Outer-arm linkage analysis was attempted on the isolated T3structure. While permethylated MSn data obviated the terminal6-antennal disaccharide linkage as a 1→4 or 1→6 linkage,conclusive confirmation of the linkage as a 1→2, 1→3 or acombination of the two was unattainable (data not shown).Negative-mode MSn was also attempted on the isolated T3species; however, spectra quality was insufficient (data notshown). Although not definitively characterized at this time,the data suggest the linkage between the terminal hex-hex di-saccharide of the 6-antennae is most likely a 1→2, 1→3 or amixture of both.

The spectral set for T3 highlights many important princi-ples to consider when interpreting data sets generatedduring permethylated NSI-MSn experimentation. For instance,the MS4 spectrum, m/z 1084.4, of Figure 6C represents acompilation of multiple isomeric fragment ions that were iso-lated and subsequently fragmented from the m/z 1302.6precursor ion. As shown in Figure 6B, two fragment ionsstemming from different regions of the precursor ion arecomposed of H4N1-[ene1OH1], m/z 1084.4. Consequently,the subsequent MS4 m/z 1084.4 product ion spectrum consists

640

Fig. 6. NSI-MSn spectra for peak (5) M5-2-AA collected fraction. Sequential disassembly of the T2 M5 positional isomer, prepared as a permethylated 2-AAderivatized sodium adducted species. The MSn pathway followed appears as inserts in the top left corner of the spectra (A–E). Only one of many isomeric partialstructures is presented as an associated cartoon. The cartoons are simply used to demonstrate product ion origin specific to the cartoon shown and a single MSn

fragmentation pathway unique to the topology shown in (A). Multiple cartoons and MSn fragmentation pathways specific to the intact topology could be shown foreach spectrum. To the far right of each spectrum, selected product ions are annotated. Please be advised that not all selected product ions correspond to the associatecartoon; however, the product ions do correspond to the compilation of isomeric precursor ions (e.g., partial structures) that are simultaneously isolated anddisassembled. Importantly, all product ions annotated and cartooned originate from the intact species cartooned in (A). Product ions highlighted in bold print wereselected for subsequent dissociation and constitute the MSn pathway.

JM Prien et al.


641Figure 6.



of isomeric, isobaric and “non-isobaric” fragment ions thatmay be unique to one, some or all the isomeric parent ions(Figure 8C). For example, the base peak, m/z 866.3, is theproduct of an m/z 1084.4 fragment ion originating from acleavage event occurring between the two terminal mannoseresidues of the 6-antennae (Supplemental data, Figure S5).Meanwhile, m/z 709.4 (H3-[

0,4A]) and m/z 737.3 (H3-[3,5A])

represent cross-ring product ions arising from a different m/z1084.4 parent ion having been previously cleaved betweenthe single 3-antennae mannose and the core β1,4 linked man-nose residues (Supplemental data, Figure S5). Hence,multiple isomeric fragments are present at each stage alongan MSn spectra set, and only by following specific product-precursor fragment ion pathways within that spectral set maya glycan structure be elucidated (Table II). This report aswell as others (Ashline et al. 2005, 2007; Hanneman et al.2006; Prien et al. 2008, 2009; Bleckmann et al. 2009) de-monstrates that permethylated NSI-MSn shows utility for denovo glycan characterization and the ability to differentiatestructural isomers.

12[C6]AA and 13[C6]AA stable isotope fluorescencederivatization of M5 and RNase BAn immediate concern surrounding the use of a purchasedstandard for analysis is whether the low abundant M5 isomersare of biologic origin or artifacts of the purification process. Asnoted above, 2-AA is readily available as a 13[C6] stable isoto-pic variant. Figure 7 depicts a RRRP chromatogram of three2-AA derivatized glycan samples: (1) 2-12[C6]AA labeled

glycans derived from RNase B highlighted in black; (2)2-13[C6]AA labeled M5 standard derived from porcine thyro-globulin highlighted in red; and (3) a sample mixture of2-12[C6]AA labeled glycans from RNase B “spiked” with7.4 pmol of 2-13[C6]AA labeled M5 standard from porcine thy-roglobulin highlighted in green. Importantly, 2-AA stableisotope analogs co-elute during reverse-phase chromatography,demonstrating no peak dispersion or chromatographic varianceas shown in Figure 7. A mass difference of 6 Da between theisotopic pair provides sufficient m/z value dispersion so thateach isotopically labeled species may be distinguished andquantified without reciprocal analog interference. Additionally,ion efficiency issues are mitigated due to internal isotopic la-beled references which ionize identical to their non-isotopiclabeled glycan counterpart. A more in-depth discussion of ap-plications using the 2-13[C6]AA variant will be presented in asubsequent publication. As Figure 7 illustrates, all three sam-ples contain peaks (1a), (5b), (6) and (7a), representing thevarious species of M5. These findings confirm that multipleM5 isomers exist in RNase B.

Figure 8A–E demonstrates negative-mode MS profiles ofvarious RRRP elution peaks of the 12[C6]AA labeled RNaseB released glycan sample. The presence of m/z 1354.6 in theMS profile of peaks (1a), (5b) and (7a) may be interpreted inmultiple ways: (1) the neutral loss from in-source fragmenta-tion of three (−484 Da), two (−324 Da) and one (−162 Da)mannose residues from the known presence of 12[C6]AA labeledM8, M7, and M6 glycans, respectively (Figure 8B–D); (2) thepresence of distinct M5 species that co-elute with each of the

642

Fig. 7. RRRP chromatogram of three 2-AA derivatized glycan samples: (1) 12[C6]-2-AA labeled glycans derived from RNase B; (2) 13[C6]-2-AA labeled M5

standard derived from porcine thyroglobulin; and (3) a sample mixture of 12[C6]-2-AA labeled glycans from RNase B “spiked” with 7.4 pmol of 13[C6]-2-AAlabeled M5 standard from porcine thyroglobulin.

Fig. 8. Negative-mode MS profile of peaks (1a), (3), (5b), (6) and (7a) from 12[C6]-2-AA labeled glycans derived from RNase B and a sample mixture of 12[C6]-2-AA derivatized glycans from RNase B “spiked” with 7.4 pmol of 13[C6]-2-AA labeled M5 standard derived from porcine thyroglobulin.

JM Prien et al.


643Figure 8.



higher-order oligomannose species; and (3) a combination ofboth in-source fragmentation and the presence of coeluting spe-cies. Subsequent analysis is required to conclusively determinethe origin ofm/z 1354.6 ion. As a consequence of in-source frag-mentation, extracted ion chromatograms of native speciesshould be used cautiously as a diagnostic tool for monitoringand quantifying glycans.

Figure 8F–J represents the MS profiles of peaks (1a), (3),(5b), (6) and (7a) of the blended RNase B and M5 isotopicallylabeled sample. The expected parent ion of the 13[C6]AA la-beled M5 standard has a mass offset of +6 Da, i.e., m/z1360.6. As the spectra demonstrate (Figure 8G–I), m/z1360.6 (M5-2-

13[C6]AA) from M5 standard co-elutes with m/z 1354.6 (M5-2-

12[C6]AA) from RNase B. Hence, m/z 1354.6from peaks (1a), (5b) and (7a) cannot be ascribed solely as in-source fragmentation of the M8, M7 and M6 isomers, respec-tively. Most likely, m/z 1354.6 in the RNase B profilerepresents the occurrence of both in-source fragmentation ofthe higher order oligomannose glycans, as well as the presenceof coeluting M5 species. Figure 8J represents the D2 M7 isomer(identified through permethylated MSn disassembly, data notshown). In this instance, m/z 1354.6 and m/z 1360.6 are absent,suggesting that in-source fragmentation did not occur, therebyproviding greater evidence for the presence of multiple M5 iso-mers eluting at various points in the chromatogram.

To determine the origin of the m/z 1354.6 ion in RNase B,all fractions containing this ion (peaks (1a), (5b), (6) and (7))were collected and subjected to permethylation. Permethyl de-rivatization “caps” the indigenous intact glycan structures,allowing differentiation from in-source fragment contaminantsof higher order oligomannose glycans by a +14 Da shift. Thepermethylated NSI-MSn results from the RNase B-collectedfractions demonstrate the presence of distinct M5 species ateach elution point (peaks (1a), (5b), (6) and (7)) (NSI-MSn

spectra and fragmentation pathways were similar to those de-scribed for the M5 standard, data not shown). Two conclusionscan be drawn from the stable isotope and permethylated NSI-MSn experimentation results: (1) multiple M5 isomers exist inbovine RNase B confirming a previous report (Prien et al.2009); and (2), the low abundant isomers of the purchasedM5 standard derived from porcine thyroglobulin are not arti-facts of the commercial purification process.

Discussion

The mission of this manuscript is to share our multi-methodapproach toward de novo glycan characterization and quantita-tion, including minor species at or below ~0.1%, consistentwith regulatory expectations. The M5 isomer species are inlow abundance (0.7% and 0.1% of the total) and exemplaryto demonstrate our solution toward the characterization of mi-nor species. A simple workflow devised around three corewell-established analytical procedures: (1) fluorescence deriv-atization; (2) online rapid resolution reversed-phase separationcoupled with negative-mode sequential mass spectrometry(RRRP-(−)-MSn); and (3) permethylation derivatization withNSI-MSn analysis provides comprehensive glycomer and mi-nor structural isomer coverage. For the most comprehensiveglycan coverage, an initial permethylated NSI-MSn pre-screen-ing step is required providing upfront identification and de

novo characterization of the species present in the entire glycanpool. The use of permethylated NSI-MSn is the critical aspectfor glycan structural characterization both for initial direct in-fusion pre-screening step and chromatographically separatedfractions. The excellent resolution and high sensitivity ofRRRP-(−)-mode MSn may then be utilized to rapidly trackdown, confirm and quantitate individual separated glycanstructures. Coeluting glycans and unresolved positional iso-mers are then distinguished through a subsequent round ofpermethylation derivatization and NSI-MSn on the isolatedfractions. The end result is an extensive highly sensitive glycanmap or library, depicting well-characterized structures assignedto a particular retention time. Once retention times are associ-ated with specific glycan structures (including structuralisomers), RRRP mapping may be used for routine analysis,and glycan structures may be assigned. Here, this feasiblitystudy demonstrates that both qualitative and quantitative dataare obtained for low abundant species using this multi-methodworkflow. With this approach, we have the potential to per-form de novo glycan characterization of biotherapeutics,including low abundant species exceeding regulatory expecta-tions. The application of this methodology to monoclonalantibodies will be the focus of forthcoming articles.

Materials and methodsMaterialsOligomannose-5 (M5) standard from porcine thyroglobulinwas purchased from Prozyme (San Leandro, CA). 2-AAand 2-AB labeling kits consisting of dimethyl sulfoxide(DMSO), glacial acetic acid (HOAc), the fluorescent reagentand sodium cyanoborohydride (NaCNBH3) were purchasedfrom QA-Bio (Palm Desert, CA). Sodium hydroxide beads(small), proteomics-grade bovine ribonuclease B (RNase B),12[C6]-aminobenzoic acid and 13[C6]-aminobenzoic acid werepurchased from Sigma-Aldrich (St. Louis, MO). Normal-phase tips (DPS-6S resin, 10 μL bed volume) were sourcedfrom PhyNexus Inc (San Jose, CA). Macro SpinColumns™were obtained from Harvard Apparatus (Holliston, MA).PNGase F was purchased from New England Biolabs(Ipswich, MA).

Enzymatic release and purification of RNase B N-glycansN-linked glycans from 1 mg of bovine RNase B were releasedenzymatically following directions from the supplier. Proteinand detergents were removed by a C18 Sep-Pak, and theflow-through fraction was desalted on porous graphitized car-bon as described previously (Hanneman et al. 2006). Purifiedglycans were dried by vacuum centrifugation.

Fluorescent labelingA 50-mg/mL solution of 2-AB or 2-AA labeling reagent wasprepared by dissolution of the solid reagent in a DMSO–HOAc solution (70% DMSO/30% HOAc, v/v). The solutionwas used to stabilize a 6-mg quantity of NaCNBH3. M5 stan-dard (10 μg) was resuspended in 7 μL of 2-AA or 2-ABlabeling solution and then incubated for 2 h at 65°C. Followingreaction completion, samples were resuspended in 143 μL ofHPLC-grade water. For stable isotopic labeling, a 30-mg/mLsolution of 12[C6]-2-AA or 13[C6]-2-AA were prepared in an

644

JM Prien et al.


acetate-borate-buffered methanol solution (4% sodium acetatetrihydrate [w/v], 2% boric acid [w/v]). A 100-μL aliquot of 30mg/mL 2-12[C6]AA or 2-13[C6]AA solution and 50 μL of 1 MNaBH3CN in tetrahydrofuran were added to enzymatic re-leased RNase B glycans or the M5 standard, mixed andincubated at 80°C for 1 h.

Sample cleanup with normal-phase PhyTipsExcess reducing-end label was removed from the reaction mix-ture using normal-phase DPA-6S PhyTips as describedpreviously (Prater et al. 2007). Briefly, PhyTips were prerinsedwith 20% acetonitrile (ACN) and reequilibrated with 96%ACN prior to loading of the diluted glycans. The excess labelwas removed by washing the loaded PhyTips four times with96% ACN and the glycans eluted with 20% ACN. Sampleswere dried by vacuum centrifugation and resuspended in 1mL of HPLC-grade water. Comparative sample mixtures of12[C6]AA labeled M5 standard and 13[C6]-AA labeled RNaseB-released glycans were combined during the normal-phaseDPA-6S PhyTip cleanup stage.

Rapid resolution reverse-phase HPLCM5 isomers were separated on an Agilent Zorbex Rapid Res-olution SB-C18 column (2.1 × 50 mm, 1.8 μm) connected toan Agilent 1100 HPLC equipped with online fluorescence de-tection. The excitation and emission wavelength parameterswere 360 nm/425 nm and 330 nm/420 nm for the 2-AA and2-AB, respectively. The mobile phases used were as follows:(A) 0.1% HOAc; (B) 5% ACN, 0.1% HOAc; and (C) 80%ACN, 0.1% HOAc. The system was operated at a constantflow rate of 333 μL/min and a temperature of 50°C. Followinginjection, reagents were eluted with a 2-min isocratic elution at30% mobile phase B. This was followed by an initial gradientstep of increasing mobile phase B at 0.55%/min over 40 minfor 2-AA labeled glycans and at 0.55%/min over 20 min for 2-AB labeled species and then a second gradient step of increas-ing mobile phase B of 1.43%/min over 7 min. Finally, thecolumn was regenerated with 100% mobile phase C for 4min and reequilibrated at initial condition for an additional 4min prior to subsequent injection.

Online electrospray ionization-MS and negative-mode MSn

detectionThe outlet of the chromatographic separation was coupled di-rectly to a linear ion trap mass spectrometer (Thermo LTQXL, San Jose, CA) equipped with an electrospray ionizationsource. The instrument was tuned for either 2-AA or 2-AB la-beled samples by infusion of a 1-pmol/μL solution of thecorresponding labeled oligomannose-5 standard in mobilephase A. The tuning solutions were infused at 30 μL/min intoa 30% mobile phase B background flow of 300 μL/min. Thecapillary temperature was set at 250°C, and the spray voltagewas −2.5 kV. For tandem MS/MS, parent ions were selectedwith an isolation window width of 5 m/z, the normalized colli-sion energy was set at 35%, and the activation was conducted atQ = 0.25 for 50 ms. For subsequent MSn stages, isolation widthwindow was set at 2 m/z, the CID set at 35% with an activationQ = 0.25 for 50 ms.

Solid-phase spin-column permethylationGlycans were permethylated according to a previously re-ported method (Kang et al. 2008), however slightlyoptimized for 2-AA and 2-AB derivatized species. Briefly,samples were resuspended in 60% DMSO, 37.2% iodo-methane and 2.8% water. To reduce partial permethylation,15 sample recycles, opposed to eight sample recycles for un-derivatized glycans, were employed.

Nanospray sequential mass spectrometrySequential mass spectra were obtained from linear ion trapmass spectrometer (Thermo LTQ XL) equipped with a TriVer-sa Nanomate™-automated nanospray ion source (Advion,Ithaca, NY). Signal averaging was accomplished through ad-justing the number of scans relative to the ion signalstrength. Isolation window width varied dependent upon po-tential interference from noninformative “isobar-like”fragments, averaging 0.5 m/z. Although a narrow isolationwindow width allows fewer ions to enter the trap, signal aver-aging over extended infusion time results in significantlyimproved signal. Collision parameters were left at default va-lues with normalized collision energy set to 35% or optimizedto values leaving a minimal parent ion peak. Activation Q wasset at 0.25 and activation time for 30 ms.

Data analysisData sets generated from MSn experiments possess an enor-mous wealth of data, and interpretation can be quitecumbersome. To help alleviate interpretation burden and vali-date structural assignments, the MSn spectra were given asinputs to gtSuite™ glycoanalytic software (Glycome Technol-ogies Inc., Portsmouth, NH), which provided spectralannotation and structural isomer identification as outputs. Neg-ative-mode MS/MS and MSn data were interpreted manually.

NomenclatureClass names have been established to represent monomerswith identical mass: H for hexose (mannose, glucose, galac-tose); F for deoxyhexose (fucose); N for HexNAc (GlcNAcand GalNAc); and S for sialic acid, NeuAc. Reducing-endmonomers are denoted in the lower case h, n, f, etc. For ex-ample, H4N2n represents a composition of four hexoses, twointernal HexNAcs and a reducing-end HexNAc.When subjected to dissociation, permethylated glycans pro-

duce distinguishable hydroxyl group (–OH) and pyranosylene(-ene) “scars” where glycosidic bond cleavage has occurred(Ashline et al. 2005). Product ions are assigned a residue countpaired with scars, which are denoted by (OH) and (ene), eachof which may be modified by a count. For example, H3-[ene1OH2] represents a hexose trimer with both one (ene) andtwo (OH) “scars”. Native glycans when subjected to dissocia-tion produce indistinguishable hydroxyl group (–OH) “scars”where glycosidic bond cleavage has occurred. As used herein,native product ions will be assigned a residue count paired withscars denoted by (OH) and (ene). (OH) scars will not be fol-lowed by a count subscript because of the inability todiscriminate between fragments caused by a single cleavageevent or multiple cleavage events. For example, H3-[eneOH] re-presents three hexose monomers with one (ene) and one (OH).

645



A popular structurally descriptive nomenclature has been es-tablished by Domon and Costello (1988). This nomenclature ismost suitably used once a structure can be confidently assignedto a product ion. Briefly, Domon and Costello nomenclaturedefines particular fragments as being of type A, B, C, X, Yor Z. Ion types B/Y and C/Z are complementary fragments re-sultant from cleavage around the glycosidic oxygen. B-typeions indicate an (ene) cleavage at the fragments reducingend, C-type ions indicate an (OH) at the reducing end, Y-typeions indicate an (OH) at the nonreducing end and Z-type ionsindicate an (ene) at the nonreducing end. For example, B/Y/Ynotation represents a fragment with one (ene) cleavage at thereducing end and two (OH) cleavages at the nonreducing end.The terms (ene) and (OH) do not imply location of the scars;Domon and Costello nomenclature is required for that. Assuch, the (ene)/(OH) notation is more appropriate for production annotation, and the B/C/Y/Z notation is more appropriateonce a structure can be confidently assigned to a product ion.

In this manuscript, we employ Domon and Costello nomen-clature for cross-ring fragmentation; however, residue count isdenoted as H, N, F, etc, instead of subscript (Domon and Cost-ello 1988). Examples are as follows: (1) H2-[

3,5A] represents a3,5A cross-ring fragment with two intact hexose monomers; and(2) H2-[

3,5AOH1] represents a 3,5A cross-ring fragment withtwo intact hexose residues and a single OH scar. Please notethat for H2-[

3,5AOH1] the OH1 descriptor does not imply hy-droxyl group location but simply the presence of an openhydroxyl “scar” on the fragment ion. While the proposed no-menclature provides appropriate identification withoutinference to specific structures, the product ions of cross-ringfragmentation are still not appropriately described. Therefore,in this manuscript, when appropriate or when description ne-cessitates, both Domon and Costello and the above newlyproposed nomenclature will be used to identify product ions.

Supplementary Data

Supplementary data for this article is available online at http://glycob.oxfordjournals.org/.

Abbreviations

2-AA 2-aminobenzoic acid; 2-AB 2-aminobenzamide; ACNacetonitrile; CID collision-induced dissociation; DMSO di-methyl sulfoxide; GlcNAc N-acetylglucosamine; H hexose;HOAc glacial acetic acid; HPAEC high pH anion exchangechromatography; LC-MS liquid chromatography-mass spec-trometry; M5 Man5GlcNAc2; Man mannose; MS massspectrometry; MS/MS tandem mass spectrometry; MSn sequen-tial mass spectrometry; N N-acetylhexosamine; n reducing-endN-acetylhexosamine; NP-HPLC normal-phase high perfor-mance liquid chromatography; NSI nanospray ionization;NSI-MSn nanospray sequential mass spectrometry; RP-HPLCreverse-phase high performance liquid chromatography; RRRPrapid resolution reversed-phase chromatography.

References

Alvarez-Manilla G, Warren NL, Abney T, Atwood J3rd, Azadi P, YorkWS, Pierce M, Orlando R. 2007. Tools for glycomics: relative quantita-

tion of glycans by isotopic permethylation using 13CH3I. Glycobiology.17:677–687.

Anumula KR. 2006. Advances in fluorescence derivatization methods for high-performance liquid chromatographic analysis of glycoprotein carbohydrates.Anal Biochem. 350:1–23.

Anumula KR, Dhume ST. 1998. High resolution and high sensitivity methodsfor oligosaccharide mapping and characterization by normal phase high per-formance liquid chromatography following derivatization with highlyfluorescent anthranilic acid. Glycobiology. 8:685–694.

Ashline D, Singh S, Hanneman A, Reinhold V. 2005. Congruent strategies forcarbohydrate sequencing. 1. Mining structural details by MSn. Anal Chem.77:6250–6262.

Ashline DJ, Lapadula AJ, Liu YH, Lin M, Grace M, Pramanik B, ReinholdVN. 2007. Carbohydrate structural isomers analyzed by sequential massspectrometry. Anal Chem. 79:3830–3842.

Bleckmann C, Geyer H, Reinhold V, Lieberoth A, Schachner M, Kleene R,Geyer R. 2009. Glycomic analysis of N-linked carbohydrate epitopes fromCD24 of mouse brain. J Proteome Res. 8:567–582.

Bruggink C, Wuhrer M, Koeleman CA, Barreto V, Liu Y, Pohl C, Ingendoh A,Hokke CH, Deelder AM. 2005. Oligosaccharide analysis by capillary-scalehigh-pH anion-exchange chromatography with on-line ion-trap massspectrometry. J Chromatogr B Analyt Technol Biomed Life Sci. 829:136–143.

Brull LP, Kovacik V, Thomas-Oates JE, Heerma W, Haverkamp J. 1998. So-dium-cationized oligosaccharides do not appear to undergo ‘internal residueloss’ rearrangement processes on tandem mass spectrometry. Rapid Com-mun Mass Spectrom. 12:1520–1532.

Chai W, Piskarev V, Lawson AM. 2001. Negative-ion electrospray mass spec-trometry of neutral underivatized oligosaccharides. Anal Chem. 73:651–657.

Chai W, Piskarev V, Lawson AM. 2002. Branching pattern and sequenceanalysis of underivatized oligosaccharides by combined MS/MS of singlyand doubly charged molecular ions in negative-ion electrospray mass spec-trometry. J Am Soc Mass Spectrom. 13:670–679.

Chai W, Piskarev VE, Mulloy B, Liu Y, Evans PG, Osborn HM, LawsonAM. 2006. Analysis of chain and blood group type and branching patternof sialylated oligosaccharides by negative ion electrospray tandem massspectrometry. Anal Chem. 78:1581–1592.

Chataigne G, Couderc F, Poinsot V. 2008. Polysaccharides analysis of sinorhi-zobial capside by on-line anion exchange chromatography with pulsedamperometric detection and mass spectrometry coupling. J Chromatogr A.1185:241–250.

Chen X, Flynn GC. 2007. Analysis of N-glycans from recombinant im-munoglobulin G by on-line reversed-phase high-performance liquidchromatography/mass spectrometry. Anal Biochem. 370:147–161.

Chen X, Liu YD, Flynn GC. 2009. The effect of Fc glycan forms on humanIgG2 antibody clearance in humans. Glycobiology. 19:240–249.

Ciucanu I, Costello CE. 2003. Elimination of oxidative degradation during theper-O-methylation of carbohydrates. J Am Chem Soc. 125:16213–16219.

Costello CE, Contado-Miller JM, Cipollo JF. 2007. A glycomics platform forthe analysis of permethylated oligosaccharide alditols. J Am Soc MassSpectrom. 18:1799–1812.

Delaney J, Vouros P. 2001. Liquid chromatography ion trap mass spectrometricanalysis of oligosaccharides using permethylated derivatives. Rapid Com-mun Mass Spectrom. 15:325–334.

Domon B, Costello CE. 1988. Structure elucidation of glycosphingolipids andgangliosides using high-performance tandem mass spectrometry. Biochemis-try. 27:1534–1543.

European Medicines Agency. 2007. Draft: Guideline on production and qual-ity control of monoclonal antibodies and related substances. http://www.emea.europa.eu/pdfs/human/bwp/15765307en.pdf.

Franz AH, Lebrilla CB. 2002. Evidence for long-range glycosyl transfer reac-tions in the gas phase. J Am Soc Mass Spectrom. 13:325–337.

Fu D, Chen L, O’Neill RA. 1994. A detailed structural characterization ofribonuclease B oligosaccharides by 1H NMR spectroscopy and mass spec-trometry. Carbohydr Res. 261:173–186.

Geyer H, Geyer R. 2006. Strategies for analysis of glycoprotein glycosylation.Biochim Biophys Acta. 1764:1853–1869.

Guile GR, Rudd PM, Wing DR, Prime SB, Dwek RA. 1996. A rapid high-resolution high-performance liquid chromatographic method for separatingglycan mixtures and analyzing oligosaccharide profiles. Anal Biochem.240:210–226.

Hanneman AJ, Rosa JC, Ashline D, Reinhold VN. 2006. Isomer and glycomercomplexities of core GlcNAcs in Caenorhabditis elegans. Glycobiology.16:874–890.

646

JM Prien et al.


Harvey DJ. 1999. Matrix-assisted laser desorption/ionization mass spectrome-try of carbohydrates. Mass Spectrom Rev. 18:349–450.

Harvey DJ. 2005. Collision-induced fragmentation of negative ions fromN-linked glycans derivatized with 2-aminobenzoic acid. J Mass Spectrom.40:642–653.

Harvey DJ. 2005. Fragmentation of negative ions from carbohydrates: part 1.Use of nitrate and other anionic adducts for the production of negative ionelectrospray spectra from N-linked carbohydrates. J Am Soc Mass Spectrom.16:622–630.

Harvey DJ. 2005. Fragmentation of negative ions from carbohydrates: part 2.Fragmentation of high-mannose N-linked glycans. J Am Soc Mass Spec-trom. 16:631–646.

Harvey DJ. 2005. Fragmentation of negative ions from carbohydrates: part 3.Fragmentation of hybrid and complex N-linked glycans. J Am Soc MassSpectrom. 16:647–659.

Harvey DJ. 2005. Halogeno-substituted 2-aminobenzoic acid derivatives fornegative ion fragmentation studies of N-linked carbohydrates. Rapid Com-mun Mass Spectrom. 19:397–400.

Harvey DJ, Mattu TS, Wormald MR, Royle L, Dwek RA, Rudd PM. 2002.“Internal residue loss”: rearrangements occurring during the fragmentationof carbohydrates derivatized at the reducing terminus. Anal Chem.74:734–740.

Harvey DJ, Royle L, Radcliffe CM, Rudd PM, Dwek RA. 2008. Structural andquantitative analysis of N-linked glycans by matrix-assisted laser desorptionionization and negative ion nanospray mass spectrometry. Anal Biochem.376:44–60.

International Conference on Harmonization of Technical Requirements forRegistration of Pharmaceuticals for Human Use. 2006. Impurities in newdrug substances Q3A(R2). http://www.ich.org/LOB/media/MEDIA422.pdf.

Kang P, Mechref Y, Klouckova I, Novotny MV. 2005. Solid-phase permethyla-tion of glycans for mass spectrometric analysis. Rapid Commun MassSpectrom. 19:3421–3428.

Kang P, Mechref Y, Novotny MV. 2008. High-throughput solid-phase per-methylation of glycans prior to mass spectrometry. Rapid Commun MassSpectrom. 22:721–734.

Karlsson NG, Schulz BL, Packer NH. 2004. Structural determination of neu-tral O-linked oligosaccharide alditols by negative ion LC-electrospray-MSn. J Am Soc Mass Spectrom. 15:659–672.

Karlsson NG, Wilson NL, Wirth HJ, Dawes P, Joshi H, Packer NH. 2004.Negative ion graphitised carbon nano-liquid chromatography/mass spec-trometry increases sensitivity for glycoprotein oligosaccharide analysis.Rapid Commun Mass Spectrom. 18:2282–2292.

Kawasaki N, Itoh S, Ohta M, Hayakawa T. 2003. Microanalysis of N-linkedoligosaccharides in a glycoprotein by capillary liquid chromatography/massspectrometry and liquid chromatography/tandem mass spectrometry. AnalBiochem. 316:15–22.

Kopp KSM, Werner RG. 1997. Lot-to-lot consistency of glycoproteins. In:Townsend RR and Hotchkiss AT, editors. Techniques in Glycobiology. BocaRoton: CRC Press. p. 475–489.

Lapadula AJ, Hatcher PJ, Hanneman AJ, Ashline DJ, Zhang H, Reinhold VN.2005. Congruent strategies for carbohydrate sequencing. 3. OSCAR: an algo-rithm for assigning oligosaccharide topology from MSn data. Anal Chem.77:6271–6279.

Lee YC. 1990. High-performance anion-exchange chromatography for carbo-hydrate analysis. Anal Biochem. 189:151–162.

Mechref Y, Novotny MV, Krishnan C. 2003. Structural characterization of oli-gosaccharides using MALDI-TOF/TOF tandem mass spectrometry. AnalChem. 75:4895–4903.

MorelleW, SlomiannyMC,DiemerH, Schaeffer C, vanDorsselaer A,MichalskiJC. 2004. Fragmentation characteristics of permethylated oligosaccharidesusing a matrix-assisted laser desorption/ionization two-stage time-of-flight(TOF/TOF) tandem mass spectrometer. Rapid Commun Mass Spectrom.18:2637–2649.

Ninonuevo M, An H, Yin H, Killeen K, Grimm R, Ward R, German B, LebrillaC. 2005. Nanoliquid chromatography-mass spectrometry of oligosacchar-ides employing graphitized carbon chromatography on microchip with ahigh-accuracy mass analyzer. Electrophoresis. 26:3641–3649.

Pabst M, Bondili JS, Stadlmann J, Mach L, Altmann F. 2007. Mass + retentiontime = structure: a strategy for the analysis of N-glycans by carbon LC-ESI-MS and its application to fibrin N-glycans. Anal Chem. 79:5051–5057.

Pfenninger A, Karas M, Finke B, Stahl B. 2002. Structural analysis of under-ivatized neutral human milk oligosaccharides in the negative ion mode bynano-electrospray MS(n) (part 1: methodology). J Am Soc Mass Spectrom.13:1331–1340.

Pfenninger A, Karas M, Finke B, Stahl B. 2002. Structural analysis of under-ivatized neutral human milk oligosaccharides in the negative ion mode bynano-electrospray MS(n) (part 2: application to isomeric mixtures). J AmSoc Mass Spectrom. 13:1341–1348.

Prater BD, Anumula KR, Hutchins JT. 2007. Automated sample preparationfacilitated by PhyNexus MEA purification system for oligosaccharide map-ping of glycoproteins. Anal Biochem. 369:202–209.

Prater BD, Connelly HM, Qin Q, Cockrill SL. 2009. High-throughput immu-noglobulin G N-glycan characterization using rapid resolution reverse-phasechromatography tandem mass spectrometry. Anal Biochem. 385:69–79.

Prien JM, Ashline DJ, Lapadula AJ, Zhang H, Reinhold VN. 2009. The highmannose glycans from bovine ribonuclease B isomer characterization by iontrap MS. J Am Soc Mass Spectrom. 20:539–556.

Prien JM, Huysentruyt LC, Ashline DJ, Lapadula AJ, Seyfried TN, ReinholdVN. 2008. Differentiating N-linked glycan structural isomers in metastaticand nonmetastatic tumor cells using sequential mass spectrometry. Glyco-biology. 18:353–366.

Reinhold VN, Reinhold BB, Costello CE. 1995. Carbohydrate molecularweight profiling, sequence, linkage, and branching data: ES-MS and CID.Anal Chem. 67:1772–1784.

Reinhold VN, Sheeley DM. 1998. Detailed characterization of carbohydratelinkage and sequence in an ion trap mass spectrometer: glycosphingolipids.Anal Biochem. 259:28–33.

Sheeley DM, Reinhold VN. 1998. Structural characterization of carbohy-drate sequence, linkage, and branching in a quadrupole Ion trap massspectrometer: neutral oligosaccharides and N-linked glycans. Anal Chem.70:3053–3059.

Siemiatkoski J, Lyubarskaya Y, Houde D, Tep S, Mhatre R. 2006. A comparisonof three techniques for quantitative carbohydrate analysis used in character-ization of therapeutic antibodies. Carbohydr Res. 341:410–419.

Sutton CW, O’Neill JA, Cottrell JS. 1994. Site-specific characterization of gly-coprotein carbohydrates by exoglycosidase digestion and laser desorptionmass spectrometry. Anal Biochem. 218:34–46.

Thayer JR, Rohrer JS, Avdalovic N, Gearing RP. 1998. Improvements to in-line desalting of oligosaccharides separated by high-pH anion exchangechromatography with pulsed amperometric detection. Anal Biochem.256:207–216.

Viseux N, de Hoffmann E, Domon B. 1997. Structural analysis of permethy-lated oligosaccharides by electrospray tandem mass spectrometry. AnalChem. 69:3193–3198.

Viseux N, de Hoffmann E, Domon B. 1998. Structural assignment of per-methylated oligosaccharide subunits using sequential tandem massspectrometry. Anal Chem. 70:4951–4959.

Weiskopf AS, Vouros P, Harvey DJ. 1997. Characterization of oligosaccharidecomposition and structure by quadrupole ion trap mass spectrometry. RapidCommun Mass Spectrom. 11:1493–1504.

Weiskopf AS, Vouros P, Harvey DJ. 1998. Electrospray ionization-ion trapmass spectrometry for structural analysis of complex N-linked glycoproteinoligosaccharides. Anal Chem. 70:4441–4447.

Wuhrer M, Koeleman CA, Deelder AM, Hokke CH. 2004. Normal-phasenanoscale liquid chromatography-mass spectrometry of underivatized oli-gosaccharides at low-femtomole sensitivity. Anal Chem. 76:833–838.

Wuhrer M, Koeleman CA, Hokke CH, Deelder AM. 2006. Mass spectrometryof proton adducts of fucosylated N-glycans: fucose transfer between anten-nae gives rise to misleading fragments. Rapid Commun Mass Spectrom.20:1747–1754.

Zaia J, Miller MJ, Seymour JL, Costello CE. 2007. The role of mobile pro-tons in negative ion CID of oligosaccharides. J Am Soc Mass Spectrom.18:952–960.

Zhao C, Xie B, Chan SY, Costello CE, O’Connor PB. 2008. Collisionally acti-vated dissociation and electron capture dissociation provide complementarystructural information for branched permethylated oligosaccharides. J Am SocMass Spectrom. 19:138–150.

Zhuang Z, Starkey JA, Mechref Y, Novotny MV, Jacobson SC. 2007. Elec-trophoretic analysis of N-glycans on microfluidic devices. Anal Chem.79:7170–7175.

647



a multi-method approach toward de novo glycan characterization: a

Documents