2 Capillary electrophoresis in the3 N-glycosylation analysis4 of biopharmaceuticals5 Andras Guttman

78 In N-glycosylation analysis of biopharmaceuticals, analytical need depends on the phase of the manufacturing process. All

9 important glycoanalysis steps are thoroughly discussed. Carbohydrate sequencing by exoglycosidase arrays isomers is described

10 in conjunction with capillary electrophoresis (CE) to identify linkage and positional. A possible automated workflow for glyco-

11 analysis based on CE is outlined.

12 ª 2013 Published by Elsevier Ltd.

Keywords: Automated workflow; Biopharmaceuticals; Biotherapeutics; Capillary electrophoresis (CE); Carbohydrate sequencing; Glycoanalysis;

High-sensitivity detection; Immunogenic epitope; N-glycosylation; Purity assessment

Ab 8-Am

cy CID,

sp lucos

H graph

M ose;

gl l)aspa

by gh-pr22

23

24licat25CE)26preh27othe28bee29ans30und31tures32eity33ore34ling35nts,36orth37aph38us o39xistin4041ng42se43tech44nsive45, inc46mo*T

E-

Trends in Analytical Chemistry, Vol. xxx, No. x, 2013 Trends

01

TRAC 14075 No. of Pages 10

15 May 2013

P

breviations: ADCC, Antibody-dependent cellular cytotoxicity; APTS,

totoxicity; CE-LIF, Capillary electrophoresis-laser-induced fluorescence;

ectrometry; ETD, Electron-transfer dissociation; Gal, Galactose; Glc, G

exNAc, N-acetylhexosamine; HILIC, Hydrophilic interaction chromato

atrix-assisted laser desorption/ionization mass spectrometry; Man, Mann

ycolylneuraminic acid; PNGase-F, Peptide-N4-(N-acetyl-b-glucosaminy

Design; RPLC, Reversed-phase liquid chromatography; UPLC, Ultrahi

1. Introduction

This article reviews the appcapillary electrophoresis (associated techniques for comN-glycosylation analysis of bitics. Progress in the field hasbecause the formation of glyctemplate driven, resulting in hthousands of possible strucspecificity and microheterogentheir characterization even mlenging. With respect to fulfilregulatory agency requiremeconsidered to be a purelymethod to liquid chromatogrand mass spectrometry (MS), than important addition to the eanalytical toolsets.

With the rapidly growibiotherapeutics market [1],high-resolution bioanalyticalare required for the compreheacterization of biotherapeuticsanalysis of post-translational

Andras Guttman*

MTA-PE Translational

Glycomics Research Group,

University of Pannonia,

H-8200 Egyetem u. 10,

Veszprem, Hungary

Institute of Analytical

Chemistry, AVCR, Brno,

Czech Republic

el.: +36 30 502 6619;

mail: a.guttman@neu.edu

65-9936/$ - see front matter ª 2013 Published by Elsevier Ltd. doi:http://dx.doi.org/1

lease cite this article in press as: A. Guttman, Trends Anal. Chem. (2013),

inopyrene-1,3,6-trisulphonic acid; CDC, Complement-dependent

Collision-induced dissociation; ESI-MS, Electrospray ionization mass

e; GlcNAc, N-acetylglucosamine; GU, Glucose unit; Hex, Hexose;

y; IMAC, Immobilized metal affinity chromatography; MALDI-MS,

NANA or Neu5Ac, N-acetylneuraminic acid; NGNA or Neu5Gc, N-

ragine amidase; PTM, Post-translational modification; QBD, Quality

essure liquid chromatography; WAX, Weak anion exchange

ion ofand

ensiverapeu-

n slowis not

reds to. Sitemakes

chal-recentCE is

ogonaly (LC)fferingg bio-

globalnsitive,niques

char-ludingdifica-

tions (PTMs), higher order structures andprotein aggregation, which are all impor-tant in understanding their behavior [2](e.g., changes in PTMs may influencehigher order structure formation, leadingto possible malfunctioning of the biother-apeutic agent). One of the most prevalentPTMs is the carbohydrate moiety attachedto the protein, which is closely monitoredduring the manufacturing process (i.e.clone selection, product development andlot release). Glycosylation is a highlydynamic PTM [3] and one can evenextend the well-known central dogmawith the glycosylation modification step asDNA fi RNA fi Polypeptide fi Glycopro-tein. Glycans protect proteins, orientbinding faces, prevent non-specific inter-actions and increase protein stability, justto mention some of their most importantroles (e.g., N-glycan shield large areas ofprotein surfaces from proteases).

Protein-glycosylation analysis, a sub-setof analytical glycobiology, provides infor-mation about the glycan structures that

0.1016/j.trac.2013.04.006 1

http://dx.doi.org/10.1016/j.trac.2013.04.006

72 -7374 d75 s76 -7778 -79 a80 e81 .828384858687888990919293949596979899

100101102103104

10.106-107,108o109y110

111

: N-liwas

Trends Trends in Analytical Chemistry, Vol. xxx, No. x, 2013

TRAC 14075 No. of Pages 10

15 May 2013

given cell types or organisms attach to biopharmaceuticals during production. The two major glycosylationtypes occur in biotherapeutics are N-linked and O-linkecarbohydrates (see examples in Fig. 1). N-linked glycanare attached to the protein backbone through a welldefined consensus sequence of Asn-X-Ser/Thr (where Xcan be any amino acid except proline) through the sidechain amine group of the asparagine residue [4]. First,14-sugar precursor is co-translationally added to thasparagine in the polypeptide chain of the target protein

Figure 1. Main glycosylation types on biotherapeutics. Upper leftlinked structures. The symbolic representation of glycan structures

The structure of this precursor is common to mosteukaryotes, and contains 3 glucose, 9 mannose, and 2N-acetylglucosamine residues [3]. A complex set ofreactions attaches this branched structure to a carriermolecule (dolichol), which is then transferred to theappropriate point on the polypeptide chain as it istranslocated into the ER lumen and processed post-translationally. O-glycans on the other hand are boundthrough the side-chain OH group of serine or threonineresidues with no consensus sequence requirement, andhave at least 8 different core types [3].

Protein glycosylation involves the interplay of severalhundred enzymes, and mutations in glycosylation pro-cessing enzymes can significantly alter the resulting su-gar structures. In addition, glycosylation is specific tocells, proteins and sites, responsive to cell-culture con-ditions and may be modulated by bioprocessing condi-tions. As such, alterations in the process may even resultin the incorporation of immunogenic epitopes (e.g.,galactose-a-1,3-galactose (a-1,3-Gal) and N-glycolyl-neuraminic acid (NGNA) [5]} {e.g., Chinese hamsterovary (CHO) cells, a commonly used host cell line in theproduction of biotherapeutics, can synthesize galactose-

2 http://www.elsevier.com/locate/trac

Please cite this article in press as: A. Guttman, Trends Anal. Chem. (20

a-1,3-galactose epitopes, as recent studies revealed [6]}Therefore, in addition to the thorough glycosylationprofile analysis during the production of biotherapeuticsidentification of potentially immunogenic epitopes is alsvery important and strongly advised by regulatoragencies [7,8].

2. Glycosylation-analysis options

nked complex; Upper right: N-linked high mannose; Lower panel: O-adapted from [54].

112Various tools are available for the analysis of carbohy-113drates. One of the most important is NMR [9], but the114amount of material this method requires is often in the115high-lg range [10]. Lectins, specific for particular116glycosylation structures, are also widely used, often in117an array format with or without antibodies (Abs) [11].118An important development was the introduction of119hydrophilic interaction LC (HILIC) of fluorophore-labeled120glycans [12]. This method, when in conjunction with121exoglycosidase-array digestion, has been automated122using a microtiter-plate-based system for analysis at a123low level of detection [13]. MS is also a widely used124technique in carbohydrate analysis [14], but, while125powerful, it may not be quantitative [15]. In one126approach, the released glycans are permethylated in127strong base solution or using a sodium hydroxide128microcolumn [16]. The permethylated derivatives have129sufficiently different hydrophobicities to accommodate130their separation by reversed-phase LC (RPLC), followed131by matrix-assisted laser desorption/ionization MS132(MALDI-MS) [17], or electrospray ionization MS (ESI-133MS) [18]. ESI-MS is applied directly or after LC separa-134tion on porous graphitized carbon columns [19] or HILIC

13), http://dx.doi.org/10.1016/j.trac.2013.04.006

135 columns [20]. Albeit, this method can be sensitive,136 interpretation of MS/MS spectra may be difficult, and137 recovery from the porous graphitized carbon stationary138 phase maybe problematic, especially for highly sialylated139 structures [21]. Please note that the single-stage MS140 mode only reveals glycan composition (i.e. the number141 of Hex, HexNAc, and Neu5Ac), and gives no information142 on linkages and positions. However, the MS/MS mode143 can utilize fragmentation types (e.g., CID, ETD, and144 photodissociation) to provide some linkage and posi-145 tional information. But, please note that labile residues146 can break off during the ionization process leading to147 false information that can be misleading during the148 de149 pr150151 fu152 du153 5154 to155 G156 ES157 pe158 C159 C160 ti161 an162 es163 di164165 dr166 th167 sy

168or169[2

1703. Capillary electrophoresis of sugars

171Fo172ys173ti174se175di176ra177an178ru179m180co181fa182h183st184m

Trends in Analytical Chemistry, Vol. xxx, No. x, 2013 Trends

TRAC 14075 No. of Pages 10

15 May 2013

P

velopment of the biopharmaceutical-manufacturingocess.Close attention should be paid to sialic acid and corecosylation residues, which are especially sensitivering ESI-MS analysis. Loss of such residues of up to

0% was reported during ESI-MS analysis in comparisonliquid-phase analytical methods (e.g., HPLC) [22].

lycan structures can also be somewhat assessed by LC/I-MS analysis of glycopeptides [23]. Another high-rformance bioanalytical method for glycan analysis is

E with laser-induced fluorescent (LIF) detection [24].E-LIF can readily distinguish both linkage and posi-onal isomers, so it became widely used for glycan

alysis in the biomedical and biotechnology fields [25],pecially in conjunction with the exoglycosidase-gestion array [26].CE can be coupled to MS for the analysis of carbohy-ates and glycopeptides for glycoform profiling of bio-erapeutics [5,27,28]. Another advantage of CE-basedstems is the option of easy multiplexing even up to 48

br(Gu[3

G

wdeMglhsu

Figure 2. Comparative separation of APTS-labeled Glc(a1 fi 4)n and Glc(b1 filinear (Glcb1 fi 4 linkage) shapes of the respective sugar oligomers.

lease cite this article in press as: A. Guttman, Trends Anal. Chem. (2013),

96 capillaries for high-throughput applications9,30].

r decades, CE has been extensively used for the anal-is of fluorophore-labeled oligosaccharides in free solu-

on [31], or gel-filled columns [32]. In both cases, theparation of the labeled sugar structures is based onfferences in their hydrodynamic volume-to-chargetio. In gel-filled capillaries, some interaction of thealyte molecules with the sieving matrix cannot beled out, as was suggested earlier [33]. To obtain CEigration-time data with high precision (<0.05% RSD),-injection of a lower bracketing standard (migratingster than any structures in the sample mixture) and aigher bracketing standard (migrating slower than anyructures in the sample mixture) is highly recom-ended. Once the migration times are normalized by the

185acketing standards, the corresponding glucose-unit186U) values can be calculated using Equation (1) and187sed for database search for possible matching structures1884]:189

Ux ¼ Gn þMT0x �MTn

MTnþ1 �MTn

ð1Þ191191

192here GUx is the GU of the unknown glycan; Gn is the193gree of polymerization of the preceding homooligomer;194T 0x is the corrected migration time of the unknown195ycan; MTn is the migration time of the preceding196omooligomer; and, MTn+1 is the migration time of the197bsequent homooligomer.

4)n ladders. The insets show the helical (Glca1 fi 4 linkage) and

http://www.elsevier.com/locate/trac 3

http://dx.doi.org/10.1016/j.trac.2013.04.006

198 d199 e200 s201 e202 ,203 e204 t

205e206207n

208

209

210-211s212with the release of sugar moieties from the biopharma-213ceutical products (both innovative and biosimilars)214s215-216e217-218-219220221222

Trends Trends in Analytical Chemistry, Vol. xxx, No. x, 2013

TRAC 14075 No. of Pages 10

15 May 2013

Fig. 2 compares the separation of a Glc(a1 fi 4)n ana Glc(b1 fi 4)n ladder, which both comprise glucoselements but with a and b linkages, respectively. Thilinkage difference caused significant migration-timshifts due to their shape differences. As the inset showsthe Glc(a1 fi 4)n ladder is helical, while thGlc(b1 fi 4)n ladder is linear, rendering differen

Table 1. Calculated glucose unit values of Glc(b1 fi 4)n homoo-ligomers (second column) based on the degree of polymerizationof Glc(a1 fi 4)n oligomers (first column)

Glucose unit

DP Glc(b1 fi 4)

1 1.0002 1.8103 2.4884 3.5235 4.8326 6.342

Figure 3. Main sample-preparation steps for capillary electrophoresis anpartitioning by ethanol precipitation; and, (C) APTS labeling.

4 http://www.elsevier.com/locate/trac

Please cite this article in press as: A. Guttman, Trends Anal. Chem. (20

hydrodynamic volumes for sugar chains with the samdegree of polymerization (DP). Numerical representationof the corresponding GU values for the Glc(b1 fi 4)oligomers is shown in Table 1.

4. Sample-preparation issues

The main sample-preparation steps for CE analysis of Nlinked glycans are shown in Fig. 3. The process start

(Fig. 3A), followed by partitioning the released sugar[e.g., by ethanol precipitation of the remaining polypeptide chain (Fig. 3B)] and fluorophore labeling of thpartitioned sugars (Fig. 3C). Glycan release usually utilizes an endoglycosydase peptide N4-(N-acetyl-b-glucosaminyl)asparagine amidase (PNGase-F).

Using the correct pH for the PNGase-F release reaction(pH 7.0) is very important. If the pH of the reactionbuffer is too high or too low, it can cause epimerization

alysis of N-linked glycans. (A) Glycan release by PNGase F; (B) sugar

13), http://dx.doi.org/10.1016/j.trac.2013.04.006

223 or224 re225226 is227 to228 si229 at230 in231232 w233 en234 di235 le236 se237 En238 fo239240 th241 ti242 th243 su244 am245 th246 ba247 ac248 re249 th250 st251 re252 ba253 on254 on255 an256 flu257 re

258se259(L260ex261u262tr263in264bi265m266dy267be

2685

269Fu270m271vi272se273ex274lin275th276to277m278ac279re280tr

Trends in Analytical Chemistry, Vol. xxx, No. x, 2013 Trends

TRAC 14075 No. of Pages 10

15 May 2013

P

loss of sialic acids, respectively, which can change thesulting glycosylation pattern [35].The preferred enzymatic deglycosylation time at 37�C,12–16 h. This enzymatic reaction can be acceleratedas fast as 1–2 h at 50�C; however, one should con-

der the possible loss of labile residues (e.g., sialic acids)that temperature (50�C), again resulting in a changeglycosylation pattern.Other options to speed up glycan release are micro-

ave-assisted deglycosylation, immobilized PNGase Fzyme reactors, or pressure-cycling technology, as

scussed in [36]. These methods can decrease the re-ase time to as short as just a couple of minutes or evenconds. Please note that other endoglycosidases (e.g.,doglycosidase H and PNGase A) can also be employed

r N-glycan release, if necessary.Once the carbohydrate structures are released forme therapeutic glycoprotein, the next sample-prepara-

on step is their labeling by a charged fluorophore. Al-ough a variety of different labeling agents have beenggested in the past [37,38], for the time being, 8-inopyrene-1,3,6-trisulphonic acid (APTS) [39] is used

e most. APTS labeling is a simple reductive amination-sed reaction using a weak-acid catalyst (e.g., aceticid or citric acid), and sodium cyanoborohydrate asducing agent in organic medium. The lower the pK of

e catalyst, the shorter is the reaction time, but, again,rong acidity may raise stability issues for labile sugar

281su282fil283st284th285ti286th287in

sidues. The main advantages of reductive amination-sed carbohydrate labeling are that the fluorophorely reacts with the reducing ends of sugars in a simplee-step reaction with good derivatization yield (>90%)d negligible structural selectivity [39]. Since only one

orophore is attached to each glycan structure, the

sulting labeled sugars are readily quantified with highinu

Table 2. Exoglycosidase-enzyme array-based carbohydrate sequencing. The lthe matrix. {Published with permission from [41]}

Enzymes/vials 1

Neuraminidase (NANase) xb-Galactosidasa (GALase) -b-N-Acetylhexosaminidase (HEXase) -a-Mannosidase (MANase) -a-Fucosidase (FUCase) -

lease cite this article in press as: A. Guttman, Trends Anal. Chem. (2013),

nsitivity using detection by LIF or light-emitting diodeED). Fluorescent labeling is accomplished with a largecess of the labeling reagent, so removal of the

nconjugated dye is important, especially when elec-okinetic injection is the way to introduce the sampleto the separation capillary, as this method causesased sample entry favoring the labeling reagent. Theost common way to remove the excess derivatizatione is by Sephadex G10 resin or normal-phase/HILICads [40].

. Carbohydrate sequencing

ll structural elucidation of glycans, including infor-ation about the linkage and the position of the indi-dual sugar residues, is accomplished by carbohydratequencing in a step-wise or array manner using specificoglycosydase enzymes with appropriate sugar andkage specificity, as shown in Table 2 [41]. In practice,e fluorophore-labeled sugar structures are subject top-down digestion and bottom-up identification. Thiseans that first one type of sugar residue (e.g., sialicid, fucose, and GlcNAc) is removed from the non-ducing end of the carbohydrate and the resultinguncated structure is analyzed by CE-LIF. Then, the nextgar-residue types are removed and the resulting pro-es analyzed again by CE-LIF, until the N-lined coreructure of GlcNAc2Man3 is obtained. At this stage, alle CE traces are compared and, based on the migration-

me shift and time changes of the individual peaks in alle traces, the entire structure can be reconstructedcluding the position and linkage information of the

288dividual sugar-building blocks. The most frequently289sed exoglycosidase enzymes are linkage-specific sialid-

ower panel depicts the cleavage spots of the individual enzymes in

2 3 4 5

x x x xx x x x- x x x- - x x- - - x

http://www.elsevier.com/locate/trac 5

http://dx.doi.org/10.1016/j.trac.2013.04.006

290 ases, galacosidases, fucosidases, hexosaminidases and291 mannosidases [29,41]. It is important to note that292 sialylated structures have higher charge states due to the293 number of sialic-acid residues in addition to the three294 negative charges of the APTS label. These extra charges295 cause faster electrophoretic migration of these species,296 resulting in possible comigration of multiple structures at297 the early migration-time regime of the electrophero-298 gram, making structural elucidation extremely chal-299 lenging and sometimes even impossible.300 To alleviate this problem, one can apply preparative301 weak anion exchange (WAX) chromatography frac-302 tionation of the sialylated glycans with the different303 charge states before the derivatization reaction with the304 charged fluorophore. This step separates sialo structures305 (e.g., mono-, di-, tri-, and tetra-) and the fractions are306 then handled as individual glycan pools (i.e. subject to307 derivatization, purification and exoglycosidase-array-308 y309 -310311 -312 -313 r314 -315

316317

318 -319 t

320characterization and validation, as shown in Fig. 4.321During all of these steps, careful analysis of the protein322and its PTMs (e.g., glycosylation) are crucial. The first323stage of this process is clone selection for glycoprotein324therapeutics, which requires high throughput that325usually involves screening hundreds of clones, including326analysis of their glycosylation profile. Understanding the327sugar-to-function relationship is already critical during328the selection of cell lines to assure that it will provide329appropriate PTMs for the required function (Quality by330Design, QBD). Glycosylation analysis should also be ap-331plied in all the following steps of biotherapeutic pro-332duction; however, the number of samples but the speed333of analysis is not then one of the most important factors.334Finally, checking for appropriate glycosylation during lot335release is the final, crucial glycoanalysis step.336Many factors contribute to alterations in glycan pro-337cessing on recombinant glycoproteins, including the338t339-340a341components), loss of cellular organelle organization (e.g.,342,343s344t345e346s347-348f349,350o351s352e

rapeu

Trends Trends in Analytical Chemistry, Vol. xxx, No. x, 2013

TRAC 14075 No. of Pages 10

15 May 2013

based glycan sequencing). This method was successfullapplied to the analysis of heavily-sialylated biopharmaceuticals (e.g., erythropoietin) [42]. Extra charges onglycans can be caused by other groups (e.g., phosphorylation), in which case immobilized metal affinity chromatography (IMAC) was successfully utilized for theipartitioning before the application of the exoglycosidasebased sequencing [43].

6. Glycan analysis during biopharmaceuticaldevelopment and production

Manufacturing of biotherapeutics includes cloning, protein expression, protein production, purification, produc

Figure 4. The manufacturing process of recombinant the

6 http://www.elsevier.com/locate/trac

Please cite this article in press as: A. Guttman, Trends Anal. Chem. (20

expression levels of the processing enzymes in the hoscell line, monosaccharide nucleotide donor levels, cellsignaling pathways (cytokines/hormones, drugs, medi

due to pH changes), mutations in genes, gene silencingoverexpression, bioprocessing environment such atemperature, and oxygen level – just to list the importanones. Having the proper analytical toolset is thereforabsolutely necessary to ensure that the product possessecorrect glycosylation for the expected biomedical activity. Having the proper glycoanalytical toolsets is ocontemporary importance, as, in a couple of yearsdozens of biotherapeutic drugs will be off patent, scompanies will start producing their biosimilar version[44]. Biosimilars are presumably produced in a sam

tic proteins where each step requires glycosylation analysis.

13), http://dx.doi.org/10.1016/j.trac.2013.04.006

353 w354 ac355 pa356 T357 ve358 pa359 al360 m361 an362363 si364 Ig365 ca366 co367 do368 co369 ce370 to371 ac372 in373 pr374 fu375 ti376 am377 ty378 pr379 en380 pr381 ti382383 pe384 pr385 to386 ar387 ch388389 fo

390A391[4

3927

393It394ch

Figure 5. Exoglycosidase digestion based a-1,3-Gal content analysis of an NCI reference standard monoclonal antibody. {Published with per-mission from [50]}.

Trends in Analytical Chemistry, Vol. xxx, No. x, 2013 Trends

TRAC 14075 No. of Pages 10

15 May 2013

P

ay as the innovator products, but usually without ex-t knowledge of the expression system and productionrameters of manufacturing of the innovator product.

hus, the glycosylation pattern of a biosimilar can bery different, changing some of the features in com-rison to the innovative compound. Of course, it can

so result in a so-called bio-better product, but, by alleans, the glycosylation pattern should be carefully

alyzed and documented.The majority of current biopharmaceuticals (and bio-

395an396im397ti398at399im400ad401m402pr403404ta405im406h407sy4081409from the National Cancer Institute (NCI) [50]. Consec-410u411be412ep413th414ar415gl416gal-containing glycan structures in this particular ref-417er418419at420tw421te422O423N

milars) are Ab therapeutics, in particular mAbs of theG1 sub-type. IgG1 possesses approximately 2–3%rbohydrate by mass, primarily attached to the highly-nserved N-glycosylation site at ASP 297 in the CH2main of the Fc region of each heavy chain [45]. Gly-sylation variability on Ab therapeutics depends on thell line and expression conditions used, possibly leadingstructural diversity with respect to their fucose, gal-

tose, sialic acid and N-acetylglucosamine content,fluencing its biological activity, physicochemicaloperties and last but not least the ADCC and CDCnctions [46]. ADCC activity, if that is the mode of ac-

on of the Ab drug, can be enhanced by decreasing theount of core fucosylated glycans at ASP 297. N-ace-

lglucosamine and mannose residues at the same siteovide ligands for Mannose Binding Protein. The pres-ce of sialic acids somewhat suppresses ADCC andovides anti-inflammatory features, while galactosyla-

on enhances CDC function.Besides the QBD considerations, other important as-cts of glycomic analysis are the determination of theesence or the absence of potential immunogenic epi-pes, even at trace levels. a1,3-gal and NGNA moietiese antigenic, so their level should be very carefullyecked throughout the entire manufacturing process.Please note that additional glycosylation sites may be

und in the hypervariable regions of the Fab portion of

lease cite this article in press as: A. Guttman, Trends Anal. Chem. (2013),

b therapeutics and should be analyzed accordingly7].

. Detection of potentially-immunogenic epitopes

has been well documented that non-human oligosac-aride motifs of galactose-a-1,3-galactose (a-1,3-Gal)d N-glycolylneuraminic acid (Neu5Gc) may triggermunogenic response in humans [48]. As glycosyla-

on is subject to cell-culture type and conditions, alter-ions in the process can result in different levels ofmunogenic sugars [49], and, since these epitopes mayversely affect the safety of biotherapeutic products,inimizing their levels during product development andoduction is very important.Galactose-a-1,3-galactose residues usually get at-ched to the non-reducing end of glycans. The highmunogenicity of a-1,3-gal is evidenced by �1% of all

uman Abs being against this epitope. Fig. 5 depicts astematic exoglycosidase-array-based approach to a-

,3-gal residue analysis of a reference standard mAb

tive exoglycosidase-digestion steps, including alpha andta galactosidases, revealed the presence of a-1,3-galitopes on the 1-6 arm (left flowchart, structure 12), one 1-3 arm (right flowchart, structure 29) and in bothms (structure 31) of the antennary structures of mAbycans. A quantitative study showed almost 10% a-1,3-

ence material.N-acetylneuraminic acid (Neu5Ac) and its hydroxyl-ed form, N-glycolylneuraminic acid (Neu5Gc), are theo major sialic acids found in mammals and typically

rminate the antennary chains of both N-glycans and-glycans via enzymatic addition by sialyl transferases.eu5Gc is not expressed in humans due to the evolu-

http://www.elsevier.com/locate/trac 7

http://dx.doi.org/10.1016/j.trac.2013.04.006

424 -425 .426 ,427428 e429 f430 s431 -432

433

434 -435 .436 f437 e438 ,439 l440 -441 -442 g443 y444 c445446447448449450451452453454455456457458

459e460y461e462r463t464-465-466467-468s469y470471472s473o474-475e476-477e478-479r480-481available databases [34].

owch

Trends Trends in Analytical Chemistry, Vol. xxx, No. x, 2013

TRAC 14075 No. of Pages 10

15 May 2013

tionary loss of the gene encoding the enzyme that converts Neu5Ac into Neu5Gc (CMP-Neu5Ac hydroxylase)Indeed, humans possess circulating Abs against Neu5Gcso glycans attached to protein therapeutics expressed incell lines capable of Neu5Gc incorporation can have thassociated immunogenic potential [51]. CE analysis oNeu5Gc is usually done at the monosaccharide-analysilevel after labeling with a non-charged fluorophore, 2aminoacridone [52].

8. Conclusions

This article reviewed the increasing role of CE with respect to N-glycosylation analysis of biotherapeuticsGlycosylation analysis is important during all the steps othe biopharmaceutical-manufacturing process. In clonselection, identification of all relevant glycan structuresincluding relative quantification of the individuacarbohydrates, is readily accomplished by CE in a highthroughput manner. CE-LIF also provides the fast turnaround time and high sensitivity that is required durinprocess development to check glycosylation consistencand to detect the presence of possibly immunogeni

Figure 6. Automated glycan analysis fl

residues. Full structural elucidation is accomplished byexoglycosydase-array treatment followed by liquid-phaseseparations or by checking critical glycosylation featureswith MS. This includes full sequence analysis, purityassessment, quantitation and identification. In formula-tion development, possible glycosylation changes can bereadily monitored by CE. The same applies to compara-bility studies and release analytics, which both requiredetection of all sugar structures with high sensitivity andaccurate quantification under good manufacturingpractice (GMP).

To accommodate the proposal of regulatory agenciesto use orthogonal separation methods for the analysis ofbiotherapeutics, CE is one of the choices, along with

8 http://www.elsevier.com/locate/trac

Please cite this article in press as: A. Guttman, Trends Anal. Chem. (20

other glycoanalytical techniques (e.g., HILIC or MS). Thorthogonality of CE-LIF and HILIC-UPLC was recentlreported in a comparative study of analyzing fluorophorlabeled IgG glycan pools, revealing that the majostructural sugar groups eluted/migrated in differenpositions with respect to their corresponding sugar-ladder standards [53]. For example, while sialylated structures eluted late in HILIC-UPLC, they migrated early inCE-LIF. However, neutral glycans migrated later in CELIF, as their charge to hydrodynamic volume ratio walower, but eluted early in HILIC-UPLC. Approximatelthe same number of glycans was identified in bothtechniques.

An automated CE-based glycan-analysis flowchart ishown in Fig. 6. This workflow can be readily applied tthe N-glycosylation analysis of the largest group of biopharmaceuticals, mAb therapeutics. The steps includpurification of the IgG molecules by protein-A partitioning, followed by PNGase-F digestion, fluorophorlabeling, sample purification/desalting and CE separation. The CE-LIF data is then analyzed and interpreted fostructural elucidation, and also compared to publicly

art. {Adapted from [55] with permission}.

482Acknowledgments483This work was supported by MTA-PE Translational484Glycomics Grant #97101 of the Hungarian Academy of485Sciences and Fulbright Research Scholar Grant #I-486174444.

487References488[1] G. Walsh, Nat. Biotechnol. 28 (2010) 917.

489[2] S.A. Berkowitz, J.R. Engen, J.R. Mazzeo, G.B. Jones, Nat. Rev. Drug

490Discov. 11 (2012) 527.

491[3] A. Varki, R. Cummings, J. Esko, H. Freeze, P. Stanley, C.R.

492Bertozzi, G.W. Hart, M.E. Etzler, Essentials of Glycobiology, Second

493Edition., Cold Spring Harbor Laboratory Press, Cold Spring

494Harbor, NY, 2009.

13), http://dx.doi.org/10.1016/j.trac.2013.04.006

495 [4] H. Lodish, A. Berk, C.A. Kaiser, S.M. Krieger, A. Bretscher, H.M.P.

496 Ploegh, Molecular Cell Biology, Sixth Edition., Macmillan, London,

497 UK, 2007.

498 [5] D. Thakur, T. Rejtar, B.L. Karger, N.J. Washburn, C.J. Bosques,

499 N.S. Gunay, Z. Shriver, G. Venkataraman, Anal. Chem. 81 (2009)

500 8900.

501 [6] C.J. Bosques, B.E. Collins, J.W. Meador 3rd, H. Sarvaiya, J.L.

502 Murphy, G. Dellorusso, D.A. Bulik, I.H. Hsu, N. Washburn, S.F.

503 Sipsey, J.R. Myette, R. Raman, Z. Shriver, R. Sasisekharan, G.

504 Venkataraman, Nat. Biotechnol. 28 (2010) 1153.

505 [7] EMA, <http://www.ema.europa.eu/docs/en_GB/docu-

506 ment_library/Scientific_guideline/2009/09/WC500003953.pdf>

507 (2005).

508 [8] S. Kozlowski, J. Woodcock, K. Midthun, R.B. Sherman, N. Engl, J.

509 Med. 365 (2011) 385.

510 [9] A.J. Petrescu, T.D. Butters, G. Reinkensmeier, S. Petrescu, F.M.

511 Platt, R.A. Dwek, M.R. Wormald, EMBO J. 16 (1997) 4302.

512 [10] A.E. Manzi, K. Norgard-Sumnicht, S. Argade, J.D. Marth, H. van

513 Halbeek, A. Varki, Glycobiology 10 (2000) 669.

514 [11] S. Chen, T. LaRoche, D. Hamelinck, D. Bergsma, D. Brenner, D.

515 Simeone, R.E. Brand, B.B. Haab, Nat. Methods 4 (2007) 437.

516 [12] P.J. Domann, A.C. Pardos-Pardos, D.L. Fernandes, D.I. Spencer,

517 C.M. Radcliffe, L. Royle, R.A. Dwek, P.M. Rudd, Proteomics 7

518 (Suppl. 1) (2007) 70.

519 [13] L. Royle, M.P. Campbell, C.M. Radcliffe, D.M. White, D.J. Harvey,

520 J.L. Abrahams, Y.G. Kim, G.W. Henry, N.A. Shadick, M.E.

521 Weinblatt, D.M. Lee, P.M. Rudd, R.A. Dwek, Anal. Biochem.

522 376 (2008) 1.

523 [14] Y. Mechref, M.V. Novotny, Chem. Rev. 102 (2002) 321.

524 [15] J. Zaia, Chem. Biol. 15 (2008) 881.

525 [16] P. Kang, Y. Mechref, I. Klouckova, M.V. Novotny, Rapid

526 Commun. Mass Spectrom. 19 (2005) 3421.

527 [17] M.V. Novotny, Y. Mechref, J. Sep. Sci. 28 (2005) 1956.

528 [18] C.E. Costello, J.M. Contado-Miller, J.F. Cipollo, J. Am. Soc. Mass

529 Spectrom. 18 (2007) 1799.

530 [19] M. Ninonuevo, H. An, H. Yin, K. Killeen, R. Grimm, R. Ward, B.

531 German, C. Lebrilla, Electrophoresis 26 (2005) 3641.

532 [20] Y. Takegawa, H. Ito, T. Keira, K. Deguchi, H. Nakagawa, S.

533 Nishimura, J. Sep. Sci. 31 (2008) 1585.

534 [21] B.L. Schulz, N.H. Packer, N.G. Karlsson, Anal. Chem. 74 (2002)

535 6088.

536 [22] D. Chelius, Fully automated N-glycosylation analysis of monoclo-

537 nal antibodies. Washington DC, 2011.

538 [23] Z. Darula, K.F. Medzihradszky, Mol. Cell. Proteomics 1 (2009) 3.

539 [24] A. Guttman, Nature (London) 380 (1996) 461.

540 [25] L.A. Gennaro, O. Salas-Solano, S. Ma, Anal. Biochem. 355 (2006)

541 249.

542 [26] S. Ma, W. Nashabeh, Anal. Chem. 71 (1999) 5185.

543 [27] R. Haselberg, G.J. de Jong, G.W. Somsen, Anal. Chem. 85 (2013)

544 2289.

545[28] Y. Liu, O. Salas-Solano, L.A. Gennaro, Anal. Chem. 81 (2009)

5466823.

547[29] W. Laroy, R. Contreras, N. Callewaert, Nat. Protoc. 1 (2006) 397.

548[30] L.R. Ruhaak, R. Hennig, C. Huhn, M. Borowiak, R.J. Dolhain,

549A.M. Deelder, E. Rapp, M. Wuhrer, J. Proteome Res. 9 (2010)

5506655.

551[31] C. Chiesa, R.A. O�Neill, Electrophoresis 15 (1994) 1132.

552[32] M. Hong, J. Sudor, M. Stefansson, M.V. Novotny, Anal. Chem. 70

553(1998) 568.

554[33] A. Guttman, T. Pritchett, Electrophoresis 16 (1995) 1906.

555[34] <http://glycobase.nibrt.ie>.

556[35] L.A. Gennaro, O. Salas-Solano, Anal. Chem. 80 (2008) 3838.

557[36] Z. Szabo, A. Guttman, B.L. Karger, Anal. Chem. 82 (2010) 2588.

558[37] L.R. Ruhaak, G. Zauner, C. Huhn, C. Bruggink, A.M. Deelder, M.

559Wuhrer, Anal. Bioanal. Chem. 397 (2010) 3457.

560[38] C. Chiesa, C. Horvath, J. Chromatogr. 645 (1993) 337.

561[39] A. Guttman, F.-T.A. Chen, R.E. Evangelista, N. Cooke, Anal.

562Biochem. 233 (1996) 234.

563[40] M. Olajos, P. Hajos, G.K. Bonn, A. Guttman, Anal. Chem. 80

564(2008) 4241.

565[41] A. Guttman, K.W. Ulfelder, J. Chromatogr. A 781 (1997) 547.

566[42] J. Bones, N. McLoughlin, M. Hilliard, K. Wynne, B.L. Karger, P.M.

567Rudd, Anal. Chem. 83 (2011) 4154.

568[43] J. Bones, S. Mittermayr, N. McLoughlin, M. Hilliard, K. Wynne,

569G.R. Johnson, J.H. Grubb, W.S. Sly, P.M. Rudd, Anal. Chem. 83

570(2011) 5344.

571[44] B.E. Erickson, Chem. Eng. News 88 (2010) 25.

572[45] P.M. Rudd, R.A. Dwek, Crit. Rev. Biochem. Mol. Biol. 32 (1997) 1.

573[46] T.S. Raju, Curr. Opin. Immunol. 20 (2008) 471.

574[47] A. Kobata, Biochim Biophys. Acta 1780 (2008) 472.

575[48] C.H. Chung, B. Mirakhur, E. Chan, Q.T. Le, J. Berlin, M. Morse,

576B.A. Murphy, S.M. Satinover, J. Hosen, D. Mauro, R.J. Slebos, Q.

577Zhou, D. Gold, T. Hatley, D.J. Hicklin, T.A. Platts-Mills, N. Engl, J.

578Med. 358 (2008) 1109.

579[49] D.F. Smith, R.D. Larsen, S. Mattox, J.B. Lowe, R.D. Cummings, J.

580Biol. Chem. 265 (1990) 6225.

581[50] Z. Szabo, A. Guttman, J. Bones, R.L. Shand, D. Meh, B.L. Karger,

582Mol. Pharm. 9 (2012) 1612.

583[51] D. Ghaderi, R.E. Taylor, V. Padler-Karavani, S. Diaz, A. Varki, Nat.

584Biotechnol. 28 (2010) 863.

585[52] Z. Szabo, J. Bones, A. Guttman, J. Glick, B.L. Karger, Anal. Chem.

58684 (2012) 7638.

587[53] S. Mittermayr, J. Bones, M. Doherty, A. Guttman, P.M. Rudd, J.

588Proteome Res. 10 (2011) 3820.

589[54] D.J. Harvey, A.H. Merry, L. Royle, M.P. Campbell, R.A. Dwek,

590P.M. Rudd, Proteomics 9 (2009) 3796.

591[55] A. Guttman, LC-GC N. Am. 30 (2012) 412.

592

Q1

Trends in Analytical Chemistry, Vol. xxx, No. x, 2013 Trends

http://www.elsevier.com/locate/trac 9

TRAC 14075 No. of Pages 10

15 May 2013

Please cite this article in press as: A. Guttman, Trends Anal. Chem. (2013), http://dx.doi.org/10.1016/j.trac.2013.04.006