maldi-tof-ms analysis of sialylated glycans and glycopeptides using 4-chloro-α-cyanocinnamic acid...

Proteomics 2012, 12, 1337–1348 1337DOI 10.1002/pmic.201100498

RESEARCH ARTICLE

MALDI-TOF-MS analysis of sialylated glycans and

glycopeptides using 4-chloro-�-cyanocinnamic acid

matrix

Maurice H. J. Selman1, Marcus Hoffmann2, Gerhild Zauner1, Liam A. McDonnell1,Crina I. A. Balog1, Erdmann Rapp2, Andre M. Deelder1 and Manfred Wuhrer1

1 Biomolecular Mass Spectrometry Unit, Department of Parasitology, Leiden University Medical Center, Leiden,The Netherlands

2 Max Planck Institute for Dynamics of Complex Technical Systems, Magdeburg, Germany

For MALDI analysis of glycans and glycopeptides, the choice of matrix is crucial in minimizingdesialylation by mass spectrometric in-source and metastable decay. Here, we evaluated thepotential of 4-chloro-�-cyanocinnamic acid (Cl-CCA) for MALDI-TOF-MS analysis of labile sia-lylated tryptic N-glycopeptides and released N- and O-glycans. Similar to DHB, but in contrast toCHCA, the Cl-CCA matrix allowed the analysis of sialylated N-glycans and glycopeptides in neg-ative ion mode MALDI-TOF-MS. Dried droplet preparations of Cl-CCA provided microcrystalswith a homogeneous spatial distribution and high shot-to-shot repeatability similar to CHCA,which simplified the automatic measurement and improved the resolution and mass accuracy.Interestingly, reflectron-positive ion mode analysis of 1-phenyl-3-methyl-5-pyrazolone (PMP)-labeled O-glycans with Cl-CCA revealed more complete profiles than with DHB and CHCA.In conclusion, we clearly demonstrate the high potential of this rationally designed matrix forglycomics and glycoproteomics.

Keywords:

Antibody / Glycoproteomics / Glycosylation / Mass spectrometry / Matrix / Post-translational modification

Received: September 21, 2011Revised: February 2, 2012

Accepted: February 12, 2012

1 Introduction

During the last two decades, MALDI-MS and ESI-MS havebecome the major soft ionization techniques widely appliedfor the analysis of a broad range of biomolecules. MALDIis most often combined with time-of-flight (TOF) mass an-alyzers and is particularly fast and easy to use. The devel-opment of faster instruments with higher mass accuracy

Correspondence: Maurice H. J. Selman, Biomolecular Mass Spec-trometry Unit, Department of Parasitology, Leiden UniversityMedical Center, PO Box 9600, 2300 RC Leiden, The NetherlandsE-mail: [email protected]: +31-71-526-6907

Abbreviations: Cl-CCA, 4-chloro--�-cyanocinnamic acid (system-atic name 3-(4-chloro-phenyl)-2-cyano-acrylic acid); HILIC, hy-drophilic interaction liquid chromatography; PMP, 1-phenyl-3-methyl-5-pyrazolone; PNGase F, N-glycosidase F; RT, room tem-perature; THAP, 2,4,6-trihydroxyacetophenone

and sensitivity, the implementation of new lasers, the ad-vent of advanced TOF/TOF instruments, as well as the intro-duction of new MALDI matrices have improved the tech-nique and widened its application range. Since over twodecades, 2,5-dihydroxybenzoic acid (DHB) and �-cyano-4-hydroxycinnamic acid (CHCA) are the “gold standard” ma-trices for the MALDI-TOF-MS analysis of peptides, gly-copeptides, and glycans [1, 2]. Other matrices frequently ap-plied include sinapinic acid, 3-hydroxypicolinic acid, 2,6-dihydroxyacetophenone, and 2,4,6-trihydroxyacetophenone(THAP) [1, 2].

MALDI-TOF-MS is a very broadly applied method for theanalysis of glycans and glycopeptides [1–3]. However, whenlabile substituents such as sialic acids are present, the anal-ysis can be complicated as these substituents may be lost byin-source and/or metastable decay. When analyzing glycans,

Colour Online: See the article online to view Figs. 2–5 in colour.

C© 2012 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.proteomics-journal.com

1338 M. H. J. Selman et al. Proteomics 2012, 12, 1337–1348

sialic acids can be stabilized by methylation of carboxylic acidresidues, or by permethylation [4–6]. In case no such stabiliza-tion measures are undertaken, the degree of desialylation isstrongly influenced by the chosen MALDI matrix. For exam-ple, CHCA as a hot matrix tends to lead to an almost completeloss of sialic acids, while DHB that is a colder matrix is lessprone to sialic acids loss and, particularly in combinationwith negative ion mode and/or linear mode MALDI-TOF-MS, allows the analysis of sialylated glycans and glycopep-tides without derivatization of the carboxylic acid groups[2, 3, 7].

In MALDI sample preparation, DHB tends to form largeneedle-shaped crystals that originate at the periphery of thetarget spot and project toward the center [1]. Crystallizationof released N-glycans with DHB (with or without sodium)generally results in more homogeneous microcrystals in thecenter of the spot whereas glycopeptides when subjected toidentical crystallization conditions are found in the large crys-tals in the periphery [1] and are detected preferably in proto-nated (positive ion mode) or deprotonated form (negative ionmode). The macrocrystalline nature of DHB often results insample spot areas where certain analytes are highly concen-trated (sweet spots), and in a very large spatial heterogeneityof mass spectra obtained from one sample spot [8–10]. More-over, MALDI-TOF-MS of macrocrystalline sample prepara-tions on linear MALDI-TOF instruments (not on orthogonalMALDI-TOF instruments) is known to result in mass spectraof rather low resolution and mass accuracy. Recrystallizationof sample spots using ethanol has been reported as a suitablemean to obtain a microcrystalline, more homogeneous gly-can sample preparation with DHB [11], but this may lead tosignal suppression. No similar procedure, however, has beenestablished for glycopeptides.

The large crystals obtained in sample preparations withDHB and the resulting large shot-to-shot variation, low massaccuracy, and resolution made us search for an alternative,microcrystalline matrix that is compatible with the analy-sis of glycans and glycopeptides bearing labile substituentssuch as sialic acids. Here, we describe the use of 4-chloro-�-cyanocinnamic acid (Cl-CCA, systematic name 3-(4-chloro-phenyl)-2-cyano-acrylic acid) for the analysis of glycans andglycopeptides in linear- and reflectron-negative ion modeMALDI-TOF-MS. Cl-CCA has recently been introduced as arationally designed MALDI matrix for proteomics [12] andhas been found to be suitable for the analysis of labilemolecules [13]. We compared Cl-CCA with conventionallyused matrices, i.e. DHB, THAP, and CHCA (see Scheme 1in supporting information for the chemical structures) forthe analysis of sialylated tryptic glycopeptides and releasedN- and O-glycans. We found Cl-CCA to be a colder matrixthan CHCA, more similar to DHB [13], allowing the detec-tion of sialylated species. Moreover, Cl-CCA preparations re-sulted in microcrystalline sample spots for glycans and gly-copeptides, making this matrix highly suitable for automatedmeasurement without compromising resolution and massaccuracy.

2 Materials and methods

2.1 IgG purification

IgGs (ca. 20 �g) were affinity captured from total humanplasma (2 �L) as described previously [3].

2.2 Reduction and alkylation

Bovine fetuin or human apo-transferrin (1 mg; Sigma-Aldrich, Steinheim, Germany) was dissolved in 200 �L50 mM ammonium bicarbonate buffer(aq) (Fluka, Steinhem,Germany) containing 10 mM DTT(aq) (Sigma-Aldrich) and re-duced at 60�C for 40 min. Cysteine alkylation was achieved byaddition of 30 �L 100 mM iodoacetamide(aq) (Sigma-Aldrich)dissolved in 50 mM ammonium bicarbonate buffer followedby 30 min incubation at room temperature (RT) in the dark.The alkylation reaction was stopped by putting the sampleunder a fluorescent lamp (gas discharge lamp) for 30 min.

2.3 Trypsin digestion

Purified plasma IgGs (ca. 20 �g) were digested with trypsin asdescribed previously [3]. For tryptic digestion of bovine fetuinor human apo-transferrin, 92 �L (0.4 mg) of the reducedand alkylated glycoprotein sample was digested overnightat 37�C with lyophilized sequencing grade modified trypsin(Promega, Madison, WI, USA) in a 1:20 (w/w) enzyme tosubstrate ratio.

2.4 In solution N-Glycosidase F release

N-glycans were released from 60 �L (0.3 mg) of the reducedand alkylated tryptic digests of bovine fetuin or human apo-transferrin by an overnight incubation at 37�C with 10 mUPNGase F (Roche Applied Science, Mannheim, Germany).

2.5 HILIC SPE purification

Ninety-six-well plate format cellulose hydrophilic interactionliquid chromatography solid-phase extraction (HILIC SPE) oftryptic IgG glycopeptides (ca. 20 �g IgG) was performed asdescribed previously [3]. Cotton HILIC micro-SPE was per-formed as described previously [14] with minor modifications.Briefly, 1 �L of tryptic glycopeptides (corresponding to ca.4.3 �g protein) or 2 �L of PNGase F released N-glycans (cor-responding to ca. 8.6 �g protein) from bovine fetuin andhuman apo-transferrin were brought to 83% ACN and ad-sorbed on a cotton HILIC micro-SPE conditioned with 83%ACN by aspirating and dispensing the sample 20 times. Theretained analytes were washed three times with 10 �L of 83%ACN containing 0.1% trifluoroacetic acid and eluted with10 �L of water.


Proteomics 2012, 12, 1337–1348 1339

2.6 Glycopeptide and glycan spotting

Glycopeptides (1 �L) derived from 220 ng IgG, 430 ng fe-tuin or transferrin, or released glycans (1 �L) derived from860 ng fetuin or transferrin were spotted either onto an MTPAnchorChip 600/384 or polished stainless steel MALDI targetplate (Bruker Daltonics, Bremen, Germany) and allowed todry by air at RT. To the dried sample spots on the AnchorChip,1 �L of DHB (5 mg/mL 50% ACN(aq); purity not defined;Bruker Daltonics) [3], THAP (1 mg/mL 50% ACN(aq); puritynot defined; Bruker Daltonics), or THAP/diammonium cit-rate (1 mg THAP and 1 mM diammonium citrate in 1 mL50% ACN(aq)) [15] was added. To the dried sample spots onthe polished steel MALDI plate, 1 �L of CHCA (5 mg/mL50% ACN(aq); purity not defined; Bruker Daltonics) [3] or Cl-CCA (5 mg/mL 70% ACN(aq); 95% purity; Bionet Research,Camelford, Cornwall, UK) was applied. Matrix-specific spot-ting and drying conditions were chosen close or identical topreviously described effective procedures [3, 12, 15].

2.7 N-glycan permethylation

Released N-glycans (derived from 1 mg fetuin or transferrin)were applied to a reversed-phase SPE cartridge (500 mgBakerbond end capped C18; Mallinckrodt Baker, Deventer,NL, USA). The flow-through and wash (5 mL water) wereapplied to a carbon cartridge (150 mg Carbograph; Grace,Lokeren, BE, USA). The retained glycans were washed withwater (5 mL), eluted with 25% aqueous ACN (5 mL), anddried by freeze-drying. The purified and desalted glycanswere permethylated as described previously [16] and recon-stituted in 20 �L ACN. Aliquots (1 �L) were spotted onto apolished steel MALDI plate, dried by air at RT, overlaid withDHB (1 �L; 10 mg/mL 50% ACN(aq); Bruker Daltonics), andanalyzed by reflectron-positive ion mode MALDI-TOF-MSas described below.

2.8 Preparation of PMP-labeled Mucin O-glycans

O-glycans were released from mucin of bovine submaxillaryglands (Sigma-Aldrich) and labeled with 1-phenyl-3-methyl-5-pyrazolone (PMP) as described previously [17]. Theyields of the reactions were checked by nano-RP-LC-MS[17] to ensure sample suitability for O-glycan profiling.From the aqueous PMP-labeled O-glycan sample, 10�L aliquots (corresponding to approximately 500 ng ofmucin starting material) were purified with a RP C18ziptip according to the manufacturers protocol (Mil-lipore, Billerica, MA), directly eluted with 2 �L DHB(5 mg/mL 50% ACN(aq)) or 3 �L Cl-CCA (5 mg/mL 50%ACN(aq); Bionet Research) onto a MALDI target plate and

analyzed by reflectron-positive ion mode MALDI-TOF-MS asdescribed below.

2.9 MALDI-TOF-MS analysis

Samples were analyzed on an Ultraflex II MALDI-TOF/TOF-MS (Bruker Daltonics) in reflectron-positive, reflectron-negative, and linear-negative ion mode. Calibration of theMALDI-TOF-MS spectra was performed with standard pep-tide mixtures from Bruker Daltonics. The samples were irra-diated by a smartbeamTM 200 Hz solid-state laser. For reflec-tron mode analysis, a 160 ns delayed extraction was appliedafter which the ions were accelerated with 25 kV, while theions were accelerated immediately without delayed extrac-tion in the linear mode. The ions were recorded betweenm/z 1300 and 4600 (N-glycans and IgG glycopeptides), m/z1300 and 8000 (fetuin and transferrin glycopeptides), m/z960 and 4500 (permethylated fetuin and transferrin glycans),or m/z 440 and 3700 (O-glycans). The ions below m/z 300(O-glycans), 500 (N-glycans and N-glycopeptides), or 800 (per-methylated N-glycans) were suppressed. Each mass spectrumwas generated by summing the ions of 1000 (O-glycans), 2000(N-glycans and N-glycopeptides), or 4000 (permethylated N-glycans) laser shots. The laser intensity was optimized to givethe best signal-to-noise (S/N) ratio for each sample. Auto-matic MALDI-TOF-MS analysis was performed as describepreviously [3]. All data processing and evaluation were per-formed with FlexAnalysis Software (Bruker Daltonics) andMicrosoft Excel, respectively. Relative intensities of the FcN-glycopeptide species derived from IgG1 and IgG2 (eightglycoforms each) were obtained by normalization to the totalIgG subclass-specific glycopeptide intensities.

2.10 Limit of detection

To evaluate the limit of detection (LOD) of MALDI-TOF-MSwith DHB and Cl-CCA, we prepared a dilution series rangingfrom 0.001 to 50 ng/�L of a tryptic digest of a monoclonalantibody (Hoffmann la Roche, Penzberg, Germany). In ad-dition, a maltoheptaose (Sigma-Aldrich) dilution series wasprepared ranging in concentration from 0.05 to 50 ng/�L.From each dilution, 1 �L was spotted onto an AnchorChiptarget plate and dried by air. To the dried sample spots, 1�L of DHB (5 mg/mL 50% ACN(aq)) [3], or 1 �L of Cl-CCA(1 mg/mL 70% ACN(aq)) was applied. Matrix-specific spottingand drying conditions were chosen identical or close to previ-ously described effective procedures [3,12]. Spots were analy-zed in reflectron-positive and reflectron-negative ion mode.

In the Fc portion of each antibody, two N-glycans are lo-cated, which may vary in glycan composition. Calculation ofthe LOD for the Fc-derived tryptic glycopeptides of a mono-clonal antibody was performed as follows:

LOD (mol) = Lowest measurable antibody amount on spot (g) ∗ 2 (Number of Fc N-glycans)

150 000 (g/mol; molecular mass antibody) ∗ Number of observed glycoforms(1)



3 Results

3.1 MALDI-TOF-MS of IgG glycopeptides

Human polyclonal IgGs purified by protein A affinity chro-matography from plasma of a healthy volunteer were digestedby trypsin and subsequently desalted and purified using cellu-lose HILIC SPE. Purified IgG glycopeptides were spotted ontoa MALDI target plate and cocrystallized with matrix (DHB,CHCA, THAP, or Cl-CCA). Initial Cl-CCA pilot experimentswere performed with commercially obtained matrix and ma-trix kindly provided by the group of Prof. Dr. M. Karas. Thematrix solution of the commercially obtained Cl-CCA wasprepared in 50%, 60%, and 70% ACN at 5 mg/mL, whilethe matrix solution of the provided Cl-CCA was prepared asdescribed previously [12]. Matrices obtained from the two dif-ferent sources gave very similar results, with the best S/Nratio being obtained for the 70% ACN preparation (data notshown).

DHB preparations showed very dense, thick crystals cover-ing the AnchorChip spot (Fig. 1A). For THAP matrix dried byvacuum, fairly homogeneous crystals were observed. Whendried at atmospheric pressure, however, THAP preparationswere heterogeneous with long crystals more similar to DHB.By contrast, CHCA and Cl-CCA preparations showed smallcrystals with a homogeneous spatial distribution. The Cl-CCApreparations occasionally showed a white crust ring of ma-trix crystals with very small, white crystals within the ring(Fig. 1B). Analysis of the ring of such spots revealed simi-lar glycopeptide profiles with a slightly lower S/N ratio andresolution compared to profiles obtained in the center. This,

however, did not affect the automatic MALDI-TOF-MS mea-surements as the ring was not sampled with the used randomwalk laser movement. For Cl-CCA preparations, intense IgGglycopeptide signals were obtained over the entire samplespot. Spots older than 2 weeks and stored on the bench stillprovided intense glycopeptide signals. When stored undervacuum in the MALDI source during 8 h, no decrease in IgGglycopeptide intensity was observed (data not shown). Samplespots stored for over 24 h in the MALDI source, however, al-lowed registration of IgG glycopeptides only in the larger notyet fully sublimated crystals of ring. The Cl-CCA and DHBpreparations required a similar laser power for threshold sig-nals, which were ca. 2–3% higher than for CHCA or 5–6%lower than for THAP (with or without diammonium citrate)preparations.

Independent of the measurement mode and MALDI ma-trix, the most abundant signals were obtained for the gly-copeptides of IgG2 followed by IgG1 and IgG4 (Fig. 2and Supporting Information Fig. S1). The observed profilesclosely resembled the reversed-phase nano-LC-ESI-MS IgGglycopeptide profiles obtained for the same sample. Theassignments of the IgG glycopeptides were confirmed byMALDI-TOF/TOF-MS/MS and nano-LC-ESI-ion trap-MS/MS analysis of the IgG1 and IgG2 G0F species providingIgG subclass-specific peptide sequence information (data notshown). In reflectron-positive ion mode MALDI-TOF-MS us-ing CHCA (Supporting Information Fig. S1A) and Cl-CCA(Supporting Information Fig. S1B) matrix, IgG glycopep-tide spectra with a good S/N ratio were obtained showingalmost exclusively neutral glycoforms with only very mi-nor signals for the sialylated species. Similarly, sialylated

Figure 1. Magnified image of tryptic IgG glycopeptides cocrystallized with DHB onto an AnchorChip MALDI plate (A) or Cl-CCA onto apolished steel MALDI plate (B). The spot formation shows a heterogeneous macrocrystalline DHB spot and a homogeneous microcrystallineCl-CCA matrix spot. Notably, Cl-CCA matrix is producing a thin white crust ring around the matrix spot.


Proteomics 2012, 12, 1337–1348 1341

Figure 2. MALDI-TOF-MS spectra of tryptic IgG Fc N-glycopeptides cocrystallized with DHB (A, C) or Cl-CCA (B, D) and measured inlinear-negative (A, B) and reflectron-negative ion mode (C, D). Profiling of IgG Fc N-glycopeptides was performed after 96-well formatcellulose-HILIC purification. Dashed arrows, IgG2 glycopeptides; solid arrows, IgG1 glycopeptides; dark square, N-acetylglucosamine;triangle, fucose; dark circle, mannose; light circle, galactose; dark diamond, N-acetylneuraminic acid; *, in-source fragment.



Figure 3. Overlay of the IgG2 G1F species from 10 reflectron-negative ion mode MALDI-TOF-MS spectra. Tryptic IgG Fc N-glycopeptidepreparations were measured automatically on a polished steel target plate with Cl-CCA (A) or an AnchorChip plate with DHB (B) orTHAP/diammonium citrate. Dark square, N-acetylglucosamine; triangle, fucose; dark circle, mannose; light circle, galactose; dark diamond,N-acetylneuraminic acid.

glycoforms were of very low relative abundance when usingCHCA matrix in reflectron-negative ion mode (SupportingInformation Fig. S1C). By contrast, DHB (Fig. 2A and C)and Cl-CCA (Fig. 2B and D) allowed detection of sialylatedglycopeptides next to the neutral glycoforms in the linear-negative (Fig. 2A and B) and reflectron-negative ion mode(Fig. 2C and D). Interestingly, an in-source decay product atm/z 2396.96 ([M-H]−) was observed at lower relative intensitywith Cl-CCA as compared to DHB. This glycopeptides speciesis formed when the linkage breaks between an antenna Glc-NAc and a core mannose of the IgG2 G0F or G1F species.The absence of this glycan species from the original IgGglycopeptide sample was confirmed by nano-reversed-phaseliquid chromatography electrospray ionization MS, indicat-ing that the observation of this glycopeptide ion in MALDI-TOF-MS is related to in-source decay (data not shown). Withthe THAP matrix, only low-quality IgG glycopeptide profilespectra with high levels of alkali salt adducts were obtainedin reflectron-positive (Supporting Information Fig. S2A) andreflectron-negative (Supporting Information Fig. S2B) ion

mode. While the presence of diammonium citrate in theTHAP matrix significantly reduced the observed alkali saltadduct levels and improved the spectra quality (SupportingInformation Fig. S2C and D), the observed S/N ratio andresolution were consistently lower compared to the other ma-trices. In an effort to improve the IgG glycopeptide spectrawith THAP/diammonium citrate, the matrix was dried un-derneath the pierced cap similar to the procedure appliedwith DHB. While this improved the observed IgG glycopep-tide signals, the S/N ratio of ca. 45 remained much lowerthan for DHB and Cl-CCA (Fig. 3).

To evaluate the performance of DHB and Cl-CCA for gly-copeptide profiling by negative ion mode MALDI-TOF-MS,the purified IgG Fc N-glycopeptide sample was spotted eighttimes on an AnchorChip and polished steel target plate andoverlaid with DHB or Cl-CCA, respectively. For both matri-ces, a single spot was measured eight times in linear- andreflectron-negative ion mode after which one spectrum wasacquired for each of the remaining replicate spots. Massspectra were processed automatically that involved internal


Proteomics 2012, 12, 1337–1348 1343

calibration, baseline subtraction, and peak picking. Withinthe spot and between the spots, the relative standard devia-tions of the normalized peak heights for the six major gly-coforms of IgG1 and IgG2 remained below 6 and 8% forthe reflectron- and linear-negative ion mode measurements,respectively.

Next, eight replicates of the above-described IgG Fc N-glycopeptide sample were prepared, spotted onto an An-chorChip (DHB) or polished steel target plate (Cl-CCA), andanalyzed by reflectron- and linear-negative ion mode MALDI-TOF-MS. This experiment was performed three times eachon different days. Within the three experiments, the intradayand interday variability was determined for IgG1 (Support-ing Information Fig. S3) and IgG2 (Supporting InformationFig. S4). Again, the relative standard deviations for the sixmajor glycoforms of IgG1 and IgG2 in the reflectron- andlinear-negative ion mode measurements remained below 6%and 8%, respectively. The homogeneous spatial distributionof the Cl-CCA microcrystals provided a high shot-to-shot re-peatability that simplified the automatic measurement andimproved the resolution and mass accuracy of the MALDI-TOF-MS measurements as demonstrated by comparing repli-cate spectra (Fig. 3).

To evaluate the LOD of MALDI-TOF-MS with DHB and Cl-CCA, a dilution series of a tryptic digest of a monoclonal anti-body was prepared. For both matrices, the dilutions were spot-ted onto an AnchorChip MALDI target plate, cocrystallizedwith DHB or Cl-CCA, and analyzed in the reflectron-positiveand reflectron-negative ion mode. The lowest measurablemonoclonal antibody amount in the reflectron-negative ionmode with DHB or Cl-CCA was 0.1 and 0.05 ng on spot, re-spectively. In the reflectron-positive ion mode with DHB orCl-CCA, the lowest measurable monoclonal antibody amountwas 0.05 and 0.01 ng on spot, respectively. Upon reflectronmode (negative and positive) analysis with DHB or Cl-CCAof the lowest measurable monoclonal antibody amount, threedifferent glycoforms (S/N ratio ≥ 3) were observed. The LODof the reflectron-negative ion mode measurements with DHBand Cl-CCA was, thereof, determined to be approximately444 and 222 amol of glycopeptides on spot, respectively. Inreflectron-positive ion mode, the LOD was found to be 222and 44 amol glycopeptides on spot for DHB and Cl-CCA,respectively. In addition, we analyzed a maltoheptaose dilu-tion series in reflectron-positive and reflectron-negative ionmode with DHB and Cl-CCA. In the positive ion mode, bothmatrices provided a similar LOD of approximately 10 fmolon spot, while in the negative ion mode, no maltoheptaosesignals could be obtained.

3.2 MALDI-TOF-MS of highly sialylated

glycopeptides

In order to evaluate the suitability of the Cl-CCA matrix fornegative ion mode MALDI-TOF-MS of glycopeptides with ahigh level of sialylation, we additionally prepared tryptic di-

gests of bovine fetuin and human apo-transferrin, selectivelypurified the glycopeptides by cotton HILIC micro-SPE [14],and spotted the sample directly onto an AnchorChip (DHB)and polished steel (Cl-CCA) MALDI target plate. MALDI-TOF-MS analysis was performed in the reflectron- and linear-negative ion mode. Fetuin glycopeptides were in additionanalyzed in the reflectron-positive ion mode. As expected,tryptic fetuin (Supporting Information Fig. S5) and transfer-rin (Supporting Information Fig. S6C and D) glycopeptidesrevealed large unfocussed metastable peaks for highly sia-lylated structures in the reflectron mode analysis with bothmatrices. In linear-negative ion mode with DHB, good signalswere obtained for the di- and tri-antennary sialylated fetuin N-glycopeptides with the most abundant signals for the Leu145–Arg159 glycopeptide cluster (small peptide moiety), followedby the Arg72–Arg103 and Val160–Arg187 glycopeptide clusters(big and intermediate-size peptide moieties, respectively;Fig. 4A). Linear-negative ion mode MALDI-TOF-MS ofCl-CCA overlaid fetuin glycopeptides revealed similar butslightly more intense profiles compared to DHB up to m/z5000, while above m/z 5000, glycopeptide intensities wererather low (Fig. 4B). Spectra of comparable quality were ob-served for human apo-transferrin overlaid with DHB (Sup-porting Information Fig. S6A) and Cl-CCA (Supporting In-formation Fig. S6B), revealing similar relative intensitiesfor the large (Gln622–Arg642) and small (Cys421–Lys433) N-glycopeptides.

3.3 MALDI-TOF-MS of released N-glycans

The performance of the Cl-CCA matrix was also evaluatedfor MALDI-TOF-MS analysis of released N-glycans. To thisend, N-glycans were released from bovine fetuin and humanapo-transferrin with PNGase F, purified by cotton HILICmicro-SPE [14], cocrystallized with DHB (AnchorChip), Cl-CCA (polished steel), THAP, or THAP/diammonium cit-rate (AnchorChip), and analyzed by linear-negative ion modeMALDI-TOF-MS.

With DHB, highly sialylated fetuin N-glycans were de-tected with a good S/N ratio, showing predominantly singlycharged deprotonated, sodiated, and potassiated species(Fig. 5A). In addition, low amounts of adducts with twosodium, two potassium, or a sodium and potassium wereobserved for the tri- and tetra-sialylated N-glycans. Spec-tra of Cl-CCA overlaid fetuin N-glycans revealed lower rela-tive intensities for the mono-sialylated fetuin glycopeptidescompared to DHB, while similar relative intensities wereobserved for the multiple sialylated glycans with similarlevels of alkali salt adducts (Fig. 5B). Moreover, the rela-tive abundance of partially sialylated species measured withCl-CCA was lower compared to DHB. With Cl-CCA prepa-rations, very intense nonglycan-related signals were ob-served at m/z 1991.1 ([M-H]−, His313–Lys333) and 2119.3([M-H]−, His313–Gly332) that were present only at low in-tensity with DHB preparations and were identified by



Figure 4. Linear-negative on mode MALDI-TOF-MS spectra of tryptic bovine fetuin glycopeptides (SwissProt entry number: P12763). Tryp-tic glycopeptides were purified by cotton HILIC micro-SPE and measured on an AnchorChip target plate with DHB (A) and a polishedsteel target plate with Cl-CCA (B). Solid arrows, small molecular-mass glycopeptide (L145CPDCPLLAPLNDSR159); dashed arrows, inter-mediate molecular-mass glycopeptide (V160VHAVEVALATFNAESNGSYLQLVEISR187); dotted arrows, large molecular-mass glycopeptide(R72PTGEVYDIEIDTLETTCHVLDPTPL ANCSVR103); dark square, N-acetylglucosamine; triangle, fucose; dark circle, mannose; light circle,galactose; dark diamond, N-acetylneuraminic acid; *, post source decay products.

MALDI-TOF/TOF-MS as fetuin peptides. THAP (SupportingInformation Fig. S7A) and THAP/diammonium citrate (Sup-porting Information Fig. S7B) cocrystallized fetuin glycansrevealed linear negative ion mode spectra of low quality thatwere dominated by alkali salt adducts. To improve the gly-can spectra, the fetuin samples were recrystallized with 60%ethanol or 60% ethanol containing 1 mM sodium acetate,revealing similar spectra as obtained prior to the recrystal-lization (data not shown).

Analysis of Cl-CCA cocrystallized transferrin N-glycans(Supporting Information Fig. S8B) revealed spectra withN-glycan profiles and alkali salt adduct levels compara-ble to those observed with DHB (Supporting InformationFig. S8A) but with a lower S/N ratio. When analyzed inthe linear- and reflectron-positive ion mode, the Cl-CCAspots with released transferrin N-glycans revealed spectrawith low-intensity signals that predominantly originated fromsodium adducts (data not shown). Transferrin glycans cocrys-tallized with THAP (Supporting Information Fig. S8C) or

THAP/diammonium citrate (Supporting Information Fig.S8D) revealed similar linear-negative ion mode glycan pro-files as obtained with DHB and Cl-CCA. While the spectrafor transferrin N-glycans with DHB and Cl-CCA were dom-inated by deprotonated glycan species, spectra of transferrinN-glycans with THAP (with and without diammonium cit-rate) showed for a big part an exchange of acidic protons forsodium. As with the fetuin N-glycans, recrystallization of thetransferrin N-glycans with 60% ethanol or 60% ethanol con-taining 1 mM sodium acetate resulted in similar spectra asobserved prior to the recrystallization (data not shown).

3.4 MALDI-TOF-MS of permethylated N-glycans

To evaluate the suitability of Cl-CCA for quantitative MALDI-TOF-MS analysis, the released fetuin and transferrin N-glycans were permethylated, cocrystallized with DHB, andanalyzed by reflectron-positive ion mode MALDI-TOF-MS.


Proteomics 2012, 12, 1337–1348 1345

Figure 5. Linear-negative ion mode MALDI-TOF-MS spectra of PNGase F released bovine fetuin N-glycans. Released N-glycans werepurified by cotton HILIC micro-SPE and cocrystallized with DHB on an AnchorChip (A) or Cl-CCA on a polished steel target plate (B) foranalysis. ∼, Fetuin peptide H313TFSGVASVESSSGEAFHVG K333; *, sodium adduct; ¥, potassium adduct; dark square, N-acetylglucosamine;triangle, fucose; dark circle, mannose; light circle, galactose; dark diamond, N-acetylneuraminic acid.

The most abundant fetuin N-glycan intensities (Support-ing Information Fig. S9A) originated from the fully sialy-lated bi- and triantennary glycans, while only low signalswere observed for the partially sialylated bi- and trianten-nary glycans. For transferrin (Supporting Information Fig.S9B), the most abundant signals originated from the fullysialylated biantennary glycans, while low signals were ob-served for the partially sialylated biantennary glycans and fullyand partially sialylated triantennary glycans. Interestingly, theobserved N-glycan profiles after permethylation were moresimilar to the negative ion mode MALDI-TOF-MS N-glycanprofiles observed with Cl-CCA than to those observed withDHB.

3.5 Comparison of PMP-labeled O-glycan profiles

using DHB and Cl-CCA

To evaluate the applicability of Cl-CCA for profiling smallglycans, PMP-labeled mucin O-glycans were cocrystallized

with DHB (AnchorChip), CHCA, or Cl-CCA (polished steel),and analyzed by reflectron-negative and positive ion modeMALDI-TOF-MS. While in reflectron-negative ion mode,we were unable to generate spectra of good quality forPMP-labeled O-glycans, reflectron-positive analysis withCl-CCA provided spectra with a higher S/N ratio as comparedto DHB (Supporting Information Fig. S10). No PMP-labeledO-glycans were observed with CHCA (data not shown). How-ever, we were unable to detect the entire PMP-labeled mucinO-glycan set that has been described previously [17,18]. As analternative attempt to improve the mucin O-glycan coverage,we cocrystallized the PMP-labeled O-glycans with Cl-CCA onan AnchorChip plate and repeated the analysis. More intensePMP-labeled O-glycan signals were obtained revealingtwo additional mucin O-glycans (m/z [1046.5+H]+ and[1062.6+H]+) corresponding to HexNAc2NeuAc1(PMP)2

and HexNAc2NeuGc1(PMP)2, respectively (HexNAc,N-acetylhexosamine; NeuAc, N-acetylneuraminic acid;NeuGc, N-glycolylneuraminic acid; Supporting InformationFig. S11).



4 Discussion

MALDI-TOF-MS analysis of sialylated glycoconjugates of-ten suffers from massive desialylation due to in-source andmetastable decay. Recently, the group of Prof. Dr. M. Karasintroduced the rationally designed MALDI matrix Cl-CCAfor proteomics [12], which showed a high potential for theanalysis of labile molecules [13]. Compared to CHCA, Cl-CCA was found to be a colder matrix that allowed more ef-ficient analyte protonation due to the lower proton affinity.Moreover, improved peptide sequence coverage, sensitivity,and detection of labile peptides has been reported [19]. Here,we have evaluated the potential of Cl-CCA for the detectionof labile sialylated glycans and glycopeptides. To this end,tryptic glycopeptides (IgG, fetuin, and transferrin), releasedN-glycans (fetuin and transferrin), and released O-glycans(bovine submaxillary gland mucin) were prepared and ana-lyzed by MALDI-TOF-MS.

In a first attempt, we compared tryptic IgG Fc N-glycopeptide MALDI-TOF-MS profiles obtained with differ-ent matrices (i.e. DHB, THAP, CHCA, and Cl-CCA) anddetection modes (i.e. reflectron-positive, reflectron-negative,and linear-negative). The observed profiles closely resembledthe reversed-phase nano-LC-ESI-MS IgG glycopeptide pro-files obtained for the same sample. Subclass-specific peptidesequence information could be obtained by fragmentation ofthe G0F species of IgG1 and IgG2, which combined with lit-erature knowledge of IgG N-glycosylation [3, 20–25] allowedstructural assignment of the detected glycoforms with goodconfidence. While IgG1 is present at a higher concentration inthe blood circulation than IgG2, the latter showed higher MSintensities possibly due to a more efficient ionization of thephenylalanine containing IgG2 glycopeptide compared to thetyrosine containing IgG1 glycopeptide. The slightly higherhydrophobicity of the peptide portion of the IgG2 glycopep-tide may also promote the efficient cocrystallization with thematrix substances. In reflectron mode, MALDI-TOF-MS sia-lylated glycopeptides are prone to in-source and metastabledecay [2,26]. Moreover, poor ionization of sialylated glycopep-tides in the positive ion mode further hampers their detec-tion. Interestingly, with DHB and Cl-CCA, we observed in-tense signals for intact sialylated IgG Fc N-glycopeptides inreflectron- and linear-negative ion mode MALDI-TOF-MS.This might be explained by a similarly efficient ionization ofneutral and acidic tryptic glycopeptides in the negative ionmode due to acidic groups of the peptide backbones [27].Moreover, the cold nature of the DHB and Cl-CCA matrices[13] minimizes analyte fragmentation that is supported bythe low intensity of the IgG2 G0F or G1F fragment ion ob-served at m/z 2396.96 ([M-H]−) or 2398.98 ([M+H]+) (Fig. 2and Supporting Information Fig. S1). Interestingly, the rela-tive abundances of these in-source decay products were evenlower with Cl-CCA than with DHB.

Human polyclonal IgG predominantly contains Fc N-glycans with one sialic acid that tends to be more stable duringMALDI ionization than multiply sialylated structures. Analy-

sis of intact monosialylated IgG glycopeptides was thereforeaccomplished by reflectron-negative ion mode MALDI-TOF-MS. In line with literature [2, 3, 7], accurate detection of in-tact highly sialylated glycopeptides with good intensity wasonly achieved with linear-negative ion mode MALDI-TOF-MS. The most intense glycopeptide signals for DHB and Cl-CCA were recorded below m/z 5000. Above m/z 5000, thesignal strength decreases that is in line with literature datafor DHB and CHCA [1].

The Cl-CCA preparations showed a good stability whenstored at atmospheric pressure but when stored at high vac-uum matrix sublimation resulted in relatively fast spot disap-pearance that is in accordance with the previously reportedhigh vacuum volatility of this matrix [19]. While this could beconsidered a disadvantage, the duration of most automatedMALDI-TOF-MS measurements remains far below 1 minper sample spot, and thus results will be obtained before anyeffect of the matrix sublimation is perceived. Furthermore,matrix contamination in the MALDI source is significantlydecreased as the sublimation process provides self cleaningof the system by the vacuum pumps.

Macrocrystalline dried droplet preparations of DHB showa distinct distribution of peptides and carbohydrates in largecrystals offering the possibility to generate relative clean car-bohydrate spectra in presence of contaminating peptides [9].By contrast, the more homogeneous distribution of com-pounds in microcrystalline preparations such as obtainedwith Cl-CCA increases the sensitivity to contaminations.Thus, resulting spectra reveal carbohydrate information withadditional peptide contamination such as observed by usfor the N-glycans released from tryptic fetuin glycopeptides(Fig. 5).

Interestingly, when compared to DHB, Cl-CCA resulted inthe registration of highly sialylated glycopeptides at relativelyhigh abundance while partially sialylated species showed lowsignal intensities. The relative intensities of the glycan speciesobserved with Cl-CCA more closely resembled the profiles ob-served for permethylated N-glycans (Supporting InformationFig. S9) and were more in-line with their previously reportedabundance on bovine fetuin [28] than those observed withDHB. This might be explained by less in-source decay ofthe fully sialylated species with Cl-CCA preparations, thoughDHB is reported to be a colder matrix than Cl-CCA [13]. Alter-natively, the differences in profiles obtained with DHB andCl-CCA may be caused by different ionization efficienciesof fully sialylated versus partially sialylated and unsialylatedstructures.

For Cl-CCA, high matrix background levels in negative ionmode MALDI-TOF-MS have already been described [13]. Forpositive ion mode measurements, this background has beenattributed to high levels of alkali salt contamination in thesample or matrix [29, 30]. The Cl-CCA matrix commerciallyobtained as a white powder with 95% purity revealed similarlevels of alkali salt adducts as the highly purified DHB matrix.Additional purification of the Cl-CCA matrix by recrystalliza-tion might further reduce the level of observed sodium and


Proteomics 2012, 12, 1337–1348 1347

potassium adducts [19]. The high level of matrix cluster back-ground with relatively low S/N ratio of released (hydrophilic)N-glycans in negative ion mode MALDI-TOF-MS is probablycaused by the low tendency of these analytes to cocrystallizewith the relative hydrophobic matrix and this matrix may,therefore, be better suited for labeled glycans. Indeed, Cl-CCA cocrystallized PMP-labeled O-glycans analyzed in posi-tive ion mode revealed superior signals as compared to DHB(see Supporting Information Figs. S10 and S11). Structuralassignment of the detected glycoforms was performed on thebasis of literature knowledge [17, 18]. While LC-MS analysesof the PMP-labeled mucin O-glycan samples gave no indi-cations of peeling (data not shown), the unassigned peakspresent in the MALDI-TOF-MS spectra indicate that thereare still some contaminants or side reaction products presentin our purified sample. Nevertheless, we here clearly demon-strated that the use of Cl-CCA has a high potential for thepositive ion mode MALDI-TOF-MS analysis of PMP-labeledO-glycans.

In conclusion, we showed that the rationally designedMALDI matrix Cl-CCA allows the sensitive analysis of in-tact labile sialylated glycopeptides by negative ion modeMALDI-TOF-MS with a low interday variability. The micro-crystalline nature of Cl-CCA provided a high shot-to-shot re-peatability, resolution, and mass accuracy, and offered thepossibility to use stainless steel MALDI target plates thatare relatively cheap and very robust. This makes Cl-CCAa matrix with high potential in high-throughput MALDI-TOF-MS profiling of glycopeptides. Moreover, we demon-strated the usefulness of Cl-CCA as a matrix for glycan anal-ysis.

The authors thank Dr Oleg I. Klychnikov for making the pho-tographs used in Fig. 1. Dr Thorsten W. Jaskolla and Profes-sor Dr Michael Karas are acknowledged for kindly providing theCl-CCA matrix used for the initial pilot experiments. Agnes L.Hipgrave Ederveen and Carolien A. M. Koeleman are acknowl-edged for preparing and measuring the permethylated fetuin andtransferrin N-glycans. The monoclonal antibody used to deter-mine the limit of detection for MALDI-TOF-MS with Cl-CCAor DHB matrix was kindly provided by Hoffmann la Roche.Maurice H.J. Selman thanks Hoffmann la Roche for financialsupport.

The authors have declared no conflict of interest.

5 References

[1] Harvey, D. J., Matrix-assisted laser desorption/ionizationmass spectrometry of carbohydrates. Mass Spectrom. Rev.1999, 18, 349–450.

[2] Harvey, D. J., Analysis of carbohydrates and glycoconju-gates by matrix-assisted laser desorption/ionization massspectrometry: an update for 2003–2004. Mass Spectrom.Rev. 2009, 28, 273–361.

[3] Selman, M. H. J., McDonnell, L. A., Palmblad, M., Ruhaak, L.R. et al., Immunoglobulin G glycopeptide profiling by matrix-assisted laser desorption ionization Fourier transform ioncyclotron resonance mass spectrometry. Anal. Chem. 2010,82, 1073–1081.

[4] Kang, P., Mechref, Y., Klouckova, I., Novotny, M. V., Solid-phase permethylation of glycans for mass spectrometricanalysis. Rapid Commun. Mass Spectrom. 2005, 19, 3421–3428.

[5] Mechref, Y., Kang, P., Novotny, M. V., Solid-phase permethy-lation for glycomic analysis. Methods Mol. Biol 2009, 534,53–64.

[6] Guillard, M., Gloerich, J., Wessels, H. J., Morava, E. et al., Au-tomated measurement of permethylated serum N-glycansby MALDI-linear ion trap mass spectrometry. Carbohydr.Res. 2009, 344, 1550–1557.

[7] Ruhaak, L. R., Huhn, C., Waterreus, W. J., de Boer, A. R.et al., Hydrophilic interaction chromatography-based high-throughput sample preparation method for N-glycan anal-ysis from total human plasma glycoproteins. Anal. Chem.2008, 80, 6119–6126.

[8] Garden, R. W., Sweedler, J. V., Heterogeneity within MALDIsamples as revealed by mass spectrometric imaging. Anal.Chem. 2000, 72, 30–36.

[9] Bouschen, W., Spengler, B., Artifacts of MALDI samplepreparation investigated by high-resolution scanning micro-probe matrix-assisted laser desorption/ionization (SMALDI)imaging mass spectrometry. Int. J. Mass Spectrom. 2007,266, 129–137.

[10] Qiao, H., Piyadasa, G., Spicer, V., Ens, W., Analyte distribu-tions in MALDI samples using MALDI imaging mass spec-trometry. Int. J. Mass Spectrom. 2009, 281, 41–51.

[11] Ruhaak, L. R., Zauner, G., Huhn, C., Bruggink, C. et al., Gly-can labeling strategies and their use in identification andquantification. Anal. Bioanal. Chem. 2010, 397, 3457–3481.

[12] Jaskolla, T. W., Lehmann, W. D., Karas, M., 4-Chloro-alpha-cyanocinnamic acid is an advanced, rationally designedMALDI matrix. Proc. Natl. Acad. Sci. USA 2008, 105, 12200–12205.

[13] Jaskolla, T., Fuchs, B., Karas, M., Schiller, J., The new ma-trix 4-chloro-alpha-cyanocinnamic acid allows the detectionof phosphatidylethanolamine chloramines by MALDI-TOFmass spectrometry. J. Am. Soc. Mass Spectrom. 2009, 20,867–874.

[14] Selman, M. H. J., Hemayatkar, M., Deelder, A. M., Wuhrer,M., Cotton HILIC SPE microtips for microscale purificationand enrichment of glycans and glycopeptides. Anal. Chem.2011, 83, 2492–2499.

[15] Oehlers, L. P., Perez, A. N., Walter, R. B., Matrix-assistedlaser desorption/ionization time-of-flight mass spectrome-try of 4-sulfophenyl isothiocyanate-derivatized peptides onAnchorChip sample supports using the sodium-tolerant ma-trix 2,4,6-trihydroxyacetophenone and diammonium citrate.Rapid Commun. Mass Spectrom. 2005, 19, 752–758.

[16] Ciucanu, I., Kerek, F., A simple and rapid method for thepermethylation of carbohydrates. Carbohydr. Res. 1984, 131,209–217.



[17] Zauner, G., Koeleman, C. A., Deelder, A. M., Wuhrer,M., Mass spectrometric O-glycan analysis after com-bined O-glycan release by beta-elimination and 1-phenyl-3-methyl-5-pyrazolone labeling. Biochim. Biophys. Acta 2011,In press.

[18] Robbe, C., Capon, C., Flahaut, C., Michalski, J. C., Mi-croscale analysis of mucin-type O-glycans by a coordi-nated fluorophore-assisted carbohydrate electrophoresisand mass spectrometry approach. Electrophoresis 2003, 24,611–621.

[19] Leszyk, J. D., Evaluation of the new MALDI matrix 4-chloro-alpha-cyanocinnamic acid. J. Biomol. Tech. 2010, 21, 81–91.

[20] Pucic, M., Knezevic, A., Vidic, J., Adamczyk, B. et al., Highthroughput isolation and glycosylation analysis of IgG –variability and heritability of the IgG glycome in three iso-lated human populations. Mol. Cell Proteomics 2011, 10,doi:10.1074/mcp.M111.010090.

[21] Wuhrer, M., Stam, J. C., van de Geijn, F. E., Koeleman, C. A.et al., Glycosylation profiling of immunoglobulin G (IgG)subclasses from human serum. Proteomics 2007, 7, 4070–4081.

[22] Stadlmann, J., Pabst, M., Kolarich, D., Kunert, R., Altmann,F., Analysis of immunoglobulin glycosylation by LC-ESI-MSof glycopeptides and oligosaccharides. Proteomics 2008, 8,2858–2871.

[23] Parekh, R. B., Dwek, R. A., Sutton, B. J., Fernandes, D. L. et al.,Association of rheumatoid arthritis and primary osteoarthri-tis with changes in the glycosylation pattern of total serumIgG. Nature 1985, 316, 452–457.

[24] Shikata, K., Yasuda, T., Takeuchi, F., Konishi, T. et al., Struc-tural changes in the oligosaccharide moiety of human IgGwith aging. Glycoconj. J. 1998, 15, 683–689.

[25] Yamada, E., Tsukamoto, Y., Sasaki, R., Yagyu, K., et al., Struc-tural changes of immunoglobulin G oligosaccharides withage in healthy human serum. Glycoconj. J. 1997, 14, 401–405.

[26] Harvey, D. J., Hunter, A. P., Bateman, R. H., Brown, J. et al.,Relationship between in-source and post-source fragmentions in the matrix-assisted laser desorption (ionization) massspectra of carbohydrates recorded with reflectron time-of-flight mass spectrometers. Int. J. Mass Spectrom. 1999, 188,131–146.

[27] Nwosu, C. C., Strum, J. S., An, H. J., Lebrilla, C. B., Enhanceddetection and identification of glycopeptides in negative ionmode mass spectrometry. Anal. Chem. 2010, 82, 9654–9662.

[28] Green, E. D., Adelt, G., Baenziger, J. U., Wilson, S. et al., Theasparagine-linked oligosaccharides on bovine fetuin. Struc-tural analysis of N-glycanase-released oligosaccharides by500-megahertz 1H NMR spectroscopy. J. Biol. Chem. 1988,263, 18253–18268.

[29] Keller, B. O., Li, L., Discerning matrix-cluster peaks inmatrix-assisted laser desorption/ionization time-of-flightmass spectra of dilute peptide mixtures. J. Am. Soc. MassSpectrom. 2000, 11, 88–93.

[30] Smirnov, I. P., Zhu, X., Taylor, T., Huang, Y. et al., Suppressionof alpha-cyano-4-hydroxycinnamic acid matrix clusters andreduction of chemical noise in MALDI-TOF mass spectrom-etry. Anal. Chem. 2004, 76, 2958–2965.


maldi-tof-ms analysis of sialylated glycans and glycopeptides using 4-chloro-α-cyanocinnamic acid...

Documents